Building Consumer Apps with ZK, MPC, and FHE | Devcon SEA

all right thanks so much Teresa um for introducing us and welcome to the panel on consumer apps um built with CK MPC and FHA I'm also super excited about this panel here um how I'm thinking of these projects is we have two of them that actually buil Primitives so ZK email and tesso they're building sort of libraries and tools and then we have two of them who built on top of them so we have ZK PDP that built on Z Emil and then cursive which uses teso as well as mpf so this is a really good representation across the stack and we'll have I have lots of questions for you guys um so I want to first start with an intro can we get um ZK email ZK PTP too and cursive in that order yeah sweet um yeah I'm aush I work a bunch on ZK email the idea is you make proofs of emails in your inbox or emails you've sent and these can get proved on chain um we've thought a lot about what are both the interesting apps you can build and what is the best kind of Def toing to do that um and on the side I also like thinking about mpcf and things like that so happy to also jump in there cool uh I'm Richard I work on ZK P2P uh we build P2P marketplaces that connect web 2 assets and uh like digital Goods with web 3 uh like smart contracts uh in particular we have like a P2P like onoff ramp uh so basically you can uh send money from VMO and use that to trade directly for usdc on chain and that's powered by ZK email um as the source of proving that you've sent that VMO payment and among other things we've also launched a couple I guess different marketplaces for trading tickets and trading domains on the secondary marketpl uh hello my name is Ash I'm from T we build uh tooling and networks to build and deploy MPC based applications um in particular coars which is what uh cursive are using and you heard a lot about this week it was the first week that it's live so we're very excited and my role there I care a lot about privacy and so I think about how using these tools in the real world actually affects the lives of the people using them cool uh my name is Andrew I work at cursive and cursive is exploring different ways for cryptography to enable new forms of human connection so specific C uh certain things that we're exploring are like multiparty FHA uh looking at karks uh with t and also like new MPC Primitives that we're like super excited about as well very cool um I want to start us off with a high level question because you guys are building in a paradigm that is very new um and a different from most consumer apps that we're used to and I'm wondering um what are some application design patterns that um are unlocked by the cryptographic Primitives and I want to address this sort of like in in like ascending order so in my mind like ZK is the like most basic primitive followed by MPC Co narts followed by MP FH so let's go in that order like what are some design patterns um that are unlocked for you and give us examples um from some of your favorite apps let's start with ZK email yeah yeah I guess we've seen two main ways people design applications that I think are different than everything we've seen before one is around proof ofed emails so that's kind of the stuff that Richard and zbp can contribute to but it's very interesting now that it's very easy to make these proofs um people are kind of starting to compose them so one example is you can prove you're part of a slack workspace by using any slack confirmation or invitation email and then also prove you for instance took a flight on United and you ended up at Thailand and now you can automatically build a system where everybody in your slack worksspace automatically gets reimbursed for their flight to Thailand and it's kind of interesting because you're not just using one proof you're combining all sorts of different proofs and just because it's very easy to build each sort of proof you can build these new applications that take multiple facets of your identity into account and so that's one that I've been really excited by by seeing um we've also seen um on the side of using emails to send transactions people kind of hot swapping almost emails and ethereum accounts in their application so for instance multi signers can just be an email instead of an ethereum address and it's kind of interesting seeing okay once that's really easy to do then people start being able to expand their apps to all sorts of new kinds of users so um I think those are the two and I think we've seen over the last few years the design evolve more into uh this being almost TurnKey for developers instead of like ZK PP which had to slog through like months of The Primitives of getting it up to speed but we learned a ton from that yeah I'm curious what your takes are though yeah I think for us at ZK PDP we're particularly focused and excited about the idea of bringing web2 data and using that like all across um blockchains especially I think if you think about like web 2 um all the data currently exists in these like centralized data silos like it's you you basically can't export them to use anywhere else think of like your Facebook follow or your Facebook friends you can't just take that and use it um on Twitter for example right and some of these like ZK Primitives allow you and I guess also like NPC with like TLS notary um allow you to kind of permissionless export uh these web2 like data from these like silos and use that anywhere uh you want and we're particularly excited about like putting these on chain uh because of this like Global like incredibly neutral settlement layer that you can do like really cool stuff and composable with who go who goes with coars you are using them I can go General Pur see well you built them I think I mean with coars and NPC for coars I think one of the nicest examples is actually what you are doing and this is essentially doing private proof delegation so with zero knowledge proofs if you want to generate a proof where um like the witness gets particularly big and you're doing it client side and maybe it's your phone is not uh powerful enough or you don't have enough time to do it on your phone you might want to Outsource it and so you see all these proof markets popping up that offer services that you know can generate your proofs for you but if you have any notion of privacy at all or if you care about about the witness data and not leaking this in the clear to everybody then you might want to do this in a privacy preserving way and I think this is where coars enter their story so doing yeah private Pro like proof delegation but maintaining privacy in an mpca I think is the one real sort of awesome use case that uh coars offer that wouldn't be possible otherwise so like in New is now you don't do proof generation on the phone you Outsource it to the network and then I think for general purpose MPC um I think the thing for me this is the most exciting part because there you start to now be able to collaborate on private data across multiple users so I think ZK is great in the sense that it gives you like this um sort of control over your data but in a way it restricts you because it's like okay I have control over my data but now it's stuck and uh I can't share it with you and sometimes I want this sort of flexibility I don't want this all or nothing and so if I have this flexibility to say okay I can share it with you and you and you or I can share something about it with you and you and you and we can compute over it together um then this is great so for me this is like the real kind of new thing that's opened up by using APC in a in a ZK world let's see yeah so I personally feel like ZK by itself is quite lame for the use cases that we're thinking about which is social applications like I feel like social is inherently like multiplayer you want to get like a bunch of people involved and the coolest applications there involve multiple people's private information and for that MPC is a perfect fit so I would say for coars the thing that's more exciting um to me personally is the idea that you can actually generate proofs about a group's data so it's like the other side of the coin and I think in general like MPC FHA all of these things uh cool applications or design patterns or like uh one thing that we're trying to explore is asking questions to your social graph asking private questions specifically so one setup here is that like Richard could like have his like private data uploaded in some like encrypted form or whatnot and I can just like ask certain queries or I can ask questions like hey like who here wants to join the hackathon with me like does anyone in my social graph meet these criteria and wants to like be part of this group um and these queries would only be decrypted if they're actually uh satisfying that criteria so I think like the MPC FHA angle feels a little more exciting for that absolutely I wanted to also so this interesting theme sort of came up about making connections so from web 2 to web 3 um from one note to another in your social gr and it's also kind of making connections between The Primitives like coars is a good example mpf is a good example so one question that's always at the the top of my mind is security and how security assumptions compose both across Primitives so what are so you are introducing new cryptographic assumptions how do these compose um and also across applications and in that case it's more about threat modeling so for example if you're relying on say the Vanos email format um and they Chang it let's say um or yeah you're um relying on like uh some dkim servers some registry like how did the threat models um compose um so I'm still curious as designers how you reason about that first of all and also how you communicate it to your users how you educate your users about it all right I'll go first um so I think think because ZK PDP uh is building like a financial application if you will uh we're dealing with like movement of I guess stable coins on chain so there is real value tied to it uh we do like treat security pretty seriously U both on the cryptographic side and on the application side um I think on the I I guess overall we see like security as kind of like how comfortable are we like securing like this dollar amount over this amount of hours for example uh and I guess like different Primitives uh that we use uh I think fall under different parts of that Spectrum uh for example like email may not be I guess the most I guess um like yeah like gated and secured uh for example whereas like behind like a gated like private API with 2fa and everything um uh yeah the data might be more secure so I think we do account for all of that in like how much we allow like the max transaction limit uh to be in some of these protocols and I think yeah I think we're constantly monitoring and seeing as these Technologies mature uh we feel more confident uh confident on it securing uh higher dollar amounts yeah I can also address trust assumptions on the ZK email side I guess we think about this in two ways one is around when things update so you say your email template updates or the de registry updates it's one very easy to change that so we have a very simple Builder where you put in the new format or the new pars thing and it automatically handles everything for you and so one thing is okay when a template changes the developer's ability to react should be extremely fast and then number two is we kind of want to make sure the system is robust to any weird kinds of like bugs or malicious DM key updates and so one way we think about this for instance with email based account recovery where it really matters like the entire security of your account hinges on you know it really shouldn't hinge on will this recovery module get hacked and so one thing we think about is like very intentional time locks and kill switches so for instance uh like many of these accounts will have uh once all the emails say hey I should recover this account to this new address a onewe time lock starts that's non-negotiable on chain and so even if either the email server got hacked or the DM key got somehow maliciously rotated or even like uh you know the users Guardians acted against their own will then at least the account owner has a week to repair that and so I think non-negotiable timel has also been quite compelling and the final thing is giving users the ability to exit so for instance for the deim keys if a user doesn't trust hey these oracles they've set for the dkim keys are are going to always act correctly we've made it so that us users can say look I have to approve all dick key rotations that are applied to my account and so now we kind of show you know in the maybe in the ux hey uh confirm your guardian emails and the relevant keys and the user just uh kind of does say some pass key authorization and that'll automatically check okay I'm going to do the DNS check did the DM Keys match okay I'm going to rotate them on chain for this user specifically so allowing users to have control over their own Keys within an intuitive manner I think is also quite important so yeah these are the three ways easy updates um having it so that there's non-negotiable time lockss and giving users control over their own keys or parts or oracles that they don't have to trust it yeah sure uh for MPC and FHA security is generally a lot more difficult like there's been so much money being put into ZK Audits and all these rollups and everything that like people know how to audit circuits very well um and for MPC fhe this is maybe becoming more of like a nent thing so we'll see how that progresses I think there are certain things you can do to make MPC FHA more secure one way to look at it is there's this notion of Direct Security versus delegated Security in NPC so delegated security is when you take your private data and you split it up and you give it to all the NPC parties and they run this computation versus Direct Security is you are actually running this computation yourself and you are one of the parties when you actually want to run compute over your own data and that's like a simple way to say oh it's your own data you should be a part of the computation and I think that simplifies part of the equation or some of the threat assumptions but I think for for MPC MPG as a whole like we're still trying to work this out maybe I can add something on like I think you talked very nicely about the composability but maybe I can try to say how this translates to users and I think to end users in particular like now we start to hear that like oh people do start to care about privacy and like okay yeah we do have stuff to hide so people do start to feel something about privacy y but I think they really don't understand like the Nuance or even the language around it so like explaining to end users like consumers that we're talking about now is just hard and like uh but even like for us now we start to kind of onboard developers to use MPC Tooling in the network and it's like even to developers it's really hard to explain and I think a lot of developers even have like a lot of confusion around the security assumptions and like it's actually a little bit worrying and I so like we and as you said like there's no auditing and like this little sort of support to for security in MPC and FHA and the slight slightly more advanced Primitives so at least we talk a lot I mean like all of us talk a lot um to each other about like okay are we right in thinking this and like with you I was talking about private proof delegation and I was saying like Okay just Outsource everything and you were like okay no we're happy to Outsource some stuff maybe but like if there's like a whistleblower or something that's you know particularly sensitive or you know then know like everything needs to be done client side so I think having conversations amongst each other to like raise the general bar of Education like e either amongst consumers and the developers and like the researchers and everyone um yeah generally just raising the level of Education I think is good and that's what we do and try to yeah inform each other I love that I think a lot of since we're building at a very early stage a lot of it is about talking to each other and communicating well communicating like in an intuitive way to users um so to that and I I want to sort of find out like some practical insights that you've gained from your experiences deploying these apps in the wild so by practical I mean like insights that you could not have gained from like just writing a white paper um just implementing it on in a hackathon for example um ux features like I use um making it clear to the users that they can control the set of De key stes um so like ux prompts how these Q behaviors in users I'm interested in that and um I'm also interested in s of how we compensate for the ux hit so a lot of these apps um they involve heavy computations right uh for example making a ZK proof for example having to be online for a long time um for the MPC or FH well not for FH but yeah um so having to wait for a long time be online for a long time like these are ux hits that we have to take how how have you been dealing with this so that's on the ux site and then I'm also interested like on the infra site like actually what what are the INF requirements like what difficulties have you run into um yeah let's go let's do the U stuff first like so like what interesting user behaviors have you observed um what sort of ux prompts ux patterns have you found useful and um how do you compensate for like the ux hit the performance hit in these apps okay um so for ZK P2P we have two flows of interacting with these like cryptographic Primitives that we use uh one the first one is obviously using ZK email um I think the main main thing about using ZK email is trying to get that email in your inbox um to yeah to start the proof generation process and that's been a difficult thing to resolve you could basically the manual way is you have to open like the raw email uh copy it and then paste it into the UI and then click a button to run the proof generation process uh so we basically built uh like a o we like got a Gmail API oath flow uh that will basically um yeah o off into your inbox uh and read the email and but uh keeping it completely local in client side uh to kick off the proof generation process um so that has helped with ux a little bit but it's obviously like a workaround kind of thing um and trying to smooth out ux things and also uh yeah after getting extracting that email the next step is obviously Genera that proof themo email circuits are like massive they like they probably will take like 8 minutes to do client side um but like 20 seconds to do on the server um so we pass it all through the server and so far like I guess all of our users are fine with that uh I think um yeah I think it it's yeah they're okay with that leaking that level of privacy um uh yeah do they know that they're sending the pl Tex email yeah yeah yeah we uh yeah we disclose everything um to to the user uh that we are doing that um yeah and then I guess the second flow I was going to talk about is we also use like TLS notary uh which uh I guess for that flow you basically need to have the users like like credentials in order to generate that like MPC proof so in order to get that you basically need to install or create a browser extension uh to scrape those like that request and response uh in in order to pass it to the NPC proof um and of course all of this is done local and client side um but yeah I think the reason for that is that's kind of non-negotiable like us like storing cookies on a server running cookies through a server and their credentials is like a complete noo whereas email is a little bit more okay uh in terms of privacy U but yeah that that's also a kind of ux hurdle we had to deal with like trying to get users to install a browser extension and going through that flow yeah maybe one thing ZK email learned a lot from zkp Top is that yeah it's always important to have the clear user flow so actually after you guys added the Google o flow and realize people are actually using this we added it to all of the existing zkm apps too where it's like you can because we noticed hey like as long as there is a way for users to do the easy path and as long as we that means we have to offer a way for the self-hosted or fully private path and so it's kind of as long as that path exists it's kind of fine to have an option for easy users who don't care to just use the kind of simple sign in with Google select the email move on path um as long as it's possible for the people who really care about privacy to get that privacy they want either they manually upload the email they select do a client side they're willing to take the ux hit in exchange for privacy or we and this is one of the advantages of having a totally open source organization is that all of our repos are open like you can go if you don't trust ZK pb's front end get the or you don't trust zk's front end go go to the GitHub repo clone it locally run it and if you really really care about privacy there's always ways to to get around the ease of use uh turn keys that we've put into the UI but for me it's kind of like we've learned okay as long as there is the easiest possible path and a way for someone who cares about privacy to do the private path we feel like that's been like a good enough mix of things even if people in general don't necessarily always do the private path um so that's one ux pattern we've noticed has been really nice um yeah yeah um bunch of thoughts here I guess first one is about privacy and like I I feel like one of the main appeals of building consumer apps with programmable cryptography is is the Privacy you get right is the choice and the consent that you have and that is like one of the major reasons why these Technologies are so powerful but I feel like if you go down the street and you just like talk to people I think what I've noticed is people just don't even understand privacy from your peers versus privacy from the server like this simple concept of like hey they look at your venmo history and if you look at it you can have like venmo or whatever payment app you use like you can have transactions only visible to your friends only visible to yourself like people think when you set your transactions to be only visible to yourself like that is privacy right that is like just privacy in general that's what that means but it's like a completely different concept when all this cryptography offers privacy from the server which is really what we're aiming for right like we don't trust server to get hacked or to be malicious or all this stuff um and I feel like this this is really like a critical question we need to solve in like communication and design of how do you actually convey this to users um so here's here's one idea I can propose which is like from from Rachel uh at cursive and also Tomo from the EF like they're both wonderful designers and they have this idea uh which they can talk more about of like hey we should have some sort of like standard UI kit for like communicating to users like when you see this interface it means it's actually private from the server like this looks totally different it feels totally different uh it is different um and that's actually connected to the technology right because right now there really is no difference and people don't understand that so I feel like having something like that where when the user actually uses the app that would maybe change their perception of things a little bit you mean like a incognito mode or something yeah yeah something like that like incognito mode like it feels different because it's like dark yeah right it's like or the I think of like the TLs like green lock icon like at the very least having some visual component where it's like oh it's not like a standard or something but it's basically this Norm you assume that if we have this green lock icon then you your connection is like more secure in some ways and we're just going to all like uphold each other to This Promise because at the end of the day like this is all up to like people building apps to just agree on and to like call out the people who don't follow this and to all work together to communicate I think that's a super nice point and like if you look at cryptographic papers or whatever when they like draw the protocols and The Primitives like they do have a sort of like it's hashed box like it's there's lines in the box when things are secret shared or it's a clear box when it's in plain text or it's black whatever so it does exist and I think this is nice but one other point that I was I wanted to mention was a bit on still on the design side and also from Rachel I guess or from cursive is like that you really play with the ideas that you know like the concepts that these programmable crypto Primitives give and you sort of yeah you play with them in your own minds and sort of somehow translate this I don't know about design or how you do it but like you somehow translate this to the user so like downstairs in the museum there's these art pieces where there's the QR code that you can scan and input an emotion and like it explains like Hey we're going to send a hash of this and like prove that you know the pre-image and like this is so cool like even Just sh showing people the language and showing people like just mentioning it just saying okay this is what's going on here I think makes people aware and like it's like okay cool I'm doing some crypto yeah and I so I was just getting lunch with Gordon who's like the artist behind that piece like basically like uh we were talking about ZK cryptography for for a while for a few hours and like he made this wonderful R piece explaining how you can keep your emotions private blah blah blah um but I feel like what what what came out of that conversation is that the most important thing for all of us who are building apps in programmable cryptography is communication like we cannot just be technologists like if we want to actually build apps that people use like we need to communicate and we need to talk to people and like explain hey this is what privacy means this is why you should want it this is what decentralization means or whatever like communication is arguably like more than half the job you can't just build the technology build The Primitives and just like throw it out there and hope something Good's going to happen but it's not even just regular communication it's like you make it look cool and fun and entertaining and like if I'm using an app I gen okay unless it's like a super serious thing that I want actual privacy like real real privacy for fine but if I'm just like using an app I'm probably doing it for some sort of entertainment or like some fun or to like engage with my friends or whatever like and then having it like look cool and feel cool and feel like a shiny cool app then it's great and I think this is something that you do well like uh it feels cool I want to be part of it so yeah I think yeah like that's what's cool about cursive is like they make it like joyful social um and like uh they sort of it's sort of turning that um privacy as like isolation and darkness kind of metaphor on its head so um that was really cool I I think yeah um all these ux insights there's no way to get them without deploying in the wild without being in production um I guess I'm also interested in sort of on the back end like um the on the infrastructure side um so I I guess I have two sort of questions about infrastructure first of all um is it expensive to run these apps so it's a leading question cuz I know it is I want to know how expensive and then like basically what improvements do you see in store what improvements are you excited about my second question is sort of a spicier one and it's about um decentralization so for example for example worldcoin um they're secret sharing their Iris codes using T um to three servers so to me three is a small number um how do we yeah I I'm wondering like another example is TLS notary so they rely on a set of trusted notaries to be the other party in the 2pc um and how big is this set so yeah it's it's kind of an infrastructure question but it's also kind of a decentralization question like um underlying these Primitives like what implementations and what infrastructure are we actually using um and yeah what are your thoughts on how this will develop should I start we laughed when you brought this up and I wanted to explain why we laughed and I don't know if I should say it like in front of a room because we haven't discussed this at all with the teams but like so like this week was the first time we deployed let's say Co snarks in like prod like properly so like this week is it's like fresh out of the oven and it's like messy it's been a bit messy it's all our fault they didn't do anything wrong it's not like so like um so so the situation is this like we have established a network so T is running a bunch of nodes cursive are running a bunch of nodes and psse are running a bunch of nodes and they all do computation they do a multi party computation between them and produce a coar and so some of the things you might learn when you go into prod that you wouldn't learn in a hackathon like as you said about like auditing and security like being immature for MPC and F and the more advanced things yeah like infrastructure and support there is super uh not in place yet so like yesterday we had um a bit of an incident where two of the nodes of one of the parties died and they had one node left and so this there our party but like okay there was some sleep involved because they stayed up all night to finish the thing to actually get it done for Devcon so it's a it's a good reason but uh yeah like one like just the nodes died and they just needed to be rebooted and um that's it but there was no like alarm or notifications or like anything set up you know this kind of like support that just keeps things online or like small stupid things that are like the absolute like complete like essential things in any sort of advanced Tech not even Advanced but like used Tech like the the most basic things are not there and we focus on just getting things working but like yeah this sort of support to keep it going it's just not there but um and in terms of cost I also don't know if I should talk about this without knowing exactly but like I think for us setting this up I think is it for each party I think it's about $7 per day for like seriously be Fe machines so yeah not too bad yeah I don't I don't have much on numbers that I know of uh concretely but I think in terms of like what you said like this is not something that we would have learned this quickly if we didn't deploy it right like this uh we were using coars for like the leaderboard for like handing out merch downstairs like if you tap like 40 people you generate a proof of it and like you can get a shirt um the shirt that I'm wearing which is super sick um but like we learned of this incident like literally 30 seconds in because people's leaderboards like weren't updating and they couldn't get their shirts and like they were so mad they were like oh where's my shirt I want my shirt like my leaderboard isn't updating so I feel like having that feedback loop instantaneously is like super useful and we like we learned this lesson and like we fix it overnight and like now we learned something new maybe I can add one good point and then I'll be Qui but uh like the good point is that we had many more we generated many many more coars than we expected so like the load was much much higher than we expected and the reason it went down was like yeah like it was not because of let's say so much load um but like we are handling much more throughput than we expected we could in like pretty good times so that's the good outcome that we wouldn't have also expected without going into prod and like a lot of it is because of the response of people at the conference actually like people are super curious and cool to try it out so that's something we wouldn't learn yeah that's a good problem like higher than expected through pit and that's a good like alert system angry users demanding shirts those are like the joys of deploying in the wild and having real users um so I do think I'm supposed to open to questions from the audience but I don't see any questions do you you don't I think questions after that time oh okay so we have 9 minutes left and then we'll open into questions great because I'm really not done here um yeah I think the second part of my question was like what do you see in the future like future improvements that you're looking forward to I in particular um want to ask about mobile interfaces and whether yeah that's a priority for you guys what challenges uh do you see like sort of what blockers do you have right now um and what challenges do you see in future or what needs to happen to make um like a joyful mobile experience for the users uh maybe one thing we've thought about on the infrastructure side and on the mobile side is um because proofs cost a couple cents on a server and will cost a couple more sents if you do coars um there is quite a pressure to push things to the client side because then it's free the user handles a computation for you and so one interesting thing is that if ZK proofs get really big and if people are doing you know not just hundreds of proofs a day but thousands or hundreds of thousands of proofs a day then there's a really big pressure economically to create client side proving I think that's quite cool because client side proving is kind of the Holy Grail of where ZK privacy comes in you don't actually leave privacy to anybody else and so I think that's it's actually potentially a good thing that it that uh we can incentivize this research um and so far there's only like you know maybe two or three organizations I can think of that are really invested in okay really we should get client side proving today because you kind of like you value in crypto often Cru to the points of centralization and by saying we're going to do client side proving you're removing that point of centralization of we run the pro we make the money um and so it's not currently economically advantageous for infrastructure providers to do client side proving but when applications are like we don't we don't want to pay you know hundreds of dollars a month or or thousands of dollars a month to you then all of a sudden applications are now incentivized to use somebody who is providing that client side proving and so um yeah that's one thing that I'm excited about in the future infrastructure wise and also for mobile proving we've seen some teams like mpro like ligron like teams in PSC like Noir really be like how can we get this client side proving done um and yeah we're we're excited to see more of that there's there's a lot of tech that has been underutilize like web GPU simd acceleration things you can get on mobile devices that you need for these email proof because people most of the time are just going to scan a QR code sign into their email on mobile and make a proof and you want to make that flow as fast as possible yeah I think for us uh we Echo a lot of the same things uh you said about client side proving for our ZK stack um I guess for us we also since we also use TLS notary and the MPC side um we're I guess we're especially also really interested in building like a mobile based flow uh to enable like yeah enable some of our marketplaces and the bottleneck there um is I guess the it's also memory where like yeah like like mobile has just much lower memory um than on I guess browser uh but yeah also like bandwidth and internet connection matters a lot in NPC um and when you're on your phone um you're typically like out um so uh you may not have access to like highspeed internet for example to like finish generating these NPC proofs and as a result like the connection drops and which is a terrible user experience uh for some people but I think there are quite a lot of like improvements like coming soon on the NPC front on the TLs notary front which uh yeah cuts the bandwidth requirement by like I don't know like like a huge like Factor um and yeah we're pretty excited to adopt that and yeah support mobile um as a Next Step yeah I think like mobile is super interesting to us just because it has such a wide toolkit for social apps like things like Bluetooth NFC like web Bluetooth web web NFC like they just suck I don't know they just don't work um or like not supported by a lot of browsers I think apple has a lot of security restrictions on on some things there um and it just sort of widens the toolkit of things that you can build uh in the social context um um so I think like mobile to me is super important and it also is just like a much more buttery experience than a than a web app but then you have the question of like how do you get people to download a mobile app I don't know yeah there's there's these things called app clips now you don't need to download a mobile app tell me about app Clips yeah it's um yeah so there's basically these app clips and instant apps on the Android store where uh like Google or um Google and Apple offer ways for you to like try out the app without actually downloading going through the App Store so you basically scan a QR code it loads some like temporary like uh like a binary on your phone and you get to interact with some limited features of the application without needing to go through the App Store maybe like one point that kind of touches on another question you wanted is um like something that I'm excited about is something something that I think both of you mentioned a little bit is about the different models of NPC so like the one where the user is actually part of the computation or when uh yeah it's a different like you delegate fully everything to the network um and so like in some cases it can be that you know as long as one per party is honest that you know the whole thing can remain secure and so if I'm involved in this protocol and I know for sure that I'm honest then I can say that this is a good like that this protocol is secure and this no collusion happening we're like quite I think not there yet at all for that but like I would love to see this happening this would be super cool oh absolutely that's like um one of the interesting thing about coars as well is because of the security guarantee from the snark site you can relax requirements on the MPC site and so yeah all all these sort of interesting like paradigms emerge um when you're like confronted with these use cases um where and they they can like these huse cases become interesting in applications like maybe in Academia they were not interesting theoretically but yeah like in practice you realize okay like this is like an opportunity um to like get some performance gains yeah so that's a really cool point I guess with my remaining two and a half minutes um I want to I want to find out what you're excited about so yeah what what applications have you seen that are like wow far out there or what applications can you imagine um what applications um would be like holy grail for you um and this can involve like uh this can be extensions of what you're working on now or it can even involve like um far out Primitives that you're looking at like witness encryption or something um yeah I'd like to hear um what what in store for the next few months and years I personally really want to see more digital signatures and more ZK poofs of those signatures I mean emails is just one example but if you can get for instance designed website standards to be more adopted you can get any website frontend now approved on chain and you can already do this with some things like Forbes in Italy um but it would be great if more websites did this I'm also excited about um pufs like physical clonable functions being the route for new open source tees I think uh like uh open source tees is kind of a holy grail and gives us the ability to to do a lot of this private computation without having to rely in uh you know Intel getting hacked every two years um what else I'm excited about witness encryption I think more research around new kinds of privacy modes and maybe I'll even that Andrew talk more about that and um yeah I'm also excited about more far out things like making certain versions of IO a little bit faster making um having NPC FHA be faster things like we've seen like Apple's w- they just released like a ML and FHA kind of application and I think seeing more of that um more Enterprises get into these privacy protocols make them faster um yeah so for me I guess um yeah like making things faster like harder better faster stronger is always the goal but I think I would love to see more like what you are doing actually and I think um um there's not so much incentive to actually play with these tools and I think the gain from playing with these tools and like onboarding users and educating them is huge but somehow people don't do it enough and so like there are apps like yours like Tonk they're building these you know like uh Salon basically where EAS Speak Easy yeah exactly where you can meet your friends like and have this private environment where you can yeah like come in if you can prove some stuff and like yeah this sort of transitive trust idea like all this kind of cool stuff um like this for me would be great I'd love to see this actually happening more and more yeah I'm super excited about new applications of MPC so specifically this concept called digital fairmon and the idea here is that like real fairmon are these chemical biochemical signals that you send around to find people you're compatible with and what if we could build the same thing with MPC but related to Digital Data digital information so you have these signals that you send around and because everything is private everything is safe just like your normal phermones are you just get to see like where your compatibility is you get to see who you match with who you might be in a group with who you might be able to do stuff with um and it just happens naturally and so one way we can actually implement this is through a new scheme uh we came up with called Trinity and Trinity is called Trinity because it's a combination of three schemes specifically kzg witness encryption uh garbled circuits and plunk uh so the rough setup here is that you have your own private data and you just upload a kzg commitment of this into the cloud or wherever and like people can just start asking you questions people can start making private queries based on this data right and essentially like you only need to respond to these queries if you actually meet those criteria like these these commitments are like fairmon from your data they're like hiding commitments to your data um and you can just like naturally have other people around you ask questions and say hey like am I compatible with this person like am I able to do something with this person uh and you can just discover that and it's only possible because MPC uh enables this privacy and safety yeah I I think I'm really excited about like more practical applications that like use these schemes that are like um being like more and more robust and secure over I guess the next year I think a lot of protocols I think ZK email went through it um and a lot of these other like protocols such as tlsn will be I guess in the next year or two and I think uh as these protocols become faster and easier to use I think we'll see more and more practical applications like ours um and yeah like stuff that you can do with web2 data uh come online and yeah I think I'm just really excited to see like combination using schemes these these different schemes in combination with each other in production yeah very cool cool um so I think now we're supposed to open to audience questions I still don't see them oh there they are so we'll start with the most uploaded which is um yeah what's the path forward for mobile do we need a new browser who will fund it I guess like we already addressed um sort of Outsourcing heavy computations from mobile whereas this question is I'm I'm interpreting it more of like um sort of the current standards of mobile like you mentioned there's taets now but there's lots of um sort of restrictions on mobile apps that make it hard to say directly interface into like TLS notary let's say um so like how do we deal with these standards that are um incompatible with The Primitives we need do we yeah how do we deal with them do we like start from scratch with a new browser what what do you guys think for mobile specifically like I feel like we just need more people building mobile apps and just experimenting like that's basically it like there's how many apps exist today on the App Store that are like uh mobile apps with programmable cryptography running clients side like I I don't I don't know we we don't have one um so I feel like we just need to learn the lessons there pretty much yeah I agree with that I think we need yeah more Dev tooling more developers uh building these kind of programmable crypto U mobile apps um I think in particular like there's a lot of pain we experienced trying to do this um in like yeah the main requirements being able to extract like the credentials um through like a web view like privately and keep it local and apple just doesn't give you or react native or any of the mobile Frameworks just don't they don't give you that tooling right now so you have to be pretty hacky um uh when you're developing I think in the mobile browser side one thing we really want is uh there's this kind of vectorization called simd and you can only access 64-bit registers in Mobile right now but if we could access access to full 256bit registers you could do ZK proofs a lot faster um similarly web GPU exists on most laptops but doesn't isn't doesn't have very good support on mobile so I think in the short term if browsers can ship those that would be super awesome and then uh but I do think the large browsers will have that eventually in Mobile within the next like year or two and that'll help accelerate a lot of these client side proofs cool um okay what are you going to have privacy by default the heart path not just privacy by complicated UI and default no privacy and limited security I don't I don't know how to parse this do you guys get the question um I guess it's kind of responding to something I used that about yeah like it's okay to present an easy path as long as we make available the fully private like hard path of like completely self-posted approving um question mark maybe one thing I can clarify on that is I I think it is possible to have decent UI for things that are private like for instance dragging and dropping the email on on like a computer is not that bad like it's still an option in the UI and it's definitely harder than sounding with Google but it's it's not like you have to download something and get clone and run locally like you can still give option to do that in the UI and we try to do that where we can and so I I do think yeah it's a fair point that foring people to self host is maybe not the best way to offer privacy but I do think you can have buttons in the UI that hey this will take a little bit longer this might require an extra action but it'll be private how far are we from ZK based transitive trust graphs of 10K plus users in DBS good question I'm looking at you I would say uh and I probably shouldn't say but I would probably say it's closer than we think I mean it's a hard problem and especially if you want to do this fully privately like suppose you want to do like a full trust graph in in MPC you need to do like essentially a full search because almost everything is connected so it's super inefficient and it's yeah not not so good now but I would imagine that like I mean the speed at which things are improving now I would say not too far away yeah I think technically it's possible for me it's like uh like how to design this like protocol Wise It's like how do you do this the right way where this is like fair for people there's a lot of like tricky I don't know moral uh product questions that are uh to be answered here MH I was also thinking just now that um fhe is one example where like you cannot offload and don't want to offload to client side you fully want the server to absorb the cost so yeah like I guess um and the tradeoffs you know um oh sorry for gesturing this is just my my opinion um like the tradeoff is um just individual users doing less work um and that could maybe allow you to scale to 10K plus users but yeah like you said there's a it's a big design space um all right how do you solve the issue of querying large dat databases efficiently in MPC all queries are linear in the database size no maybe one quick thing is like apple w like their new F system solves this by one like bucketing so you say look we're going to make square root of n buckets and we're going to send it to the right bucket then you leak a little bit of information which is which bucket did the query go to so they send square root of n queries and then it's like well you know which which query is the right query you don't know and the user throws away all about one of the queries so I thought that was a kind of clever way to you know leak a little bit of privacy but then undo that leakage uh and have it be a bit more efficient oh okay so it's like linear over square root and buckets and you can keep kind of recursively doing the square root of n uh I see yeah how about I know world does have this um they do a database search per new Iris code so yeah we do actually search over the full database for this and like it is quite optimized and it's quite efficient but like it's yeah heavily specialized optimized protocol so far um but I mean there is work definitely on um like like lookup tables and you know like there are ways I mean there people are looking into this and coming up with like clever tricks like this um but yeah so far it's uh people are looking into it but it is still an issue we need more P research PR is really cool private information retrieval isn't P also like linear work on the server site though yeah yeah yeah right now I think they you just like encrypt the whole database yeah but not for the user yeah um cool so I think we are over time now um and I really love this panel thank you so much to the panelists thank you thank you also Fork you guys I actually had to say