Episode 275 - Building in Public with Sandy & Ye from Scroll

Transcript

[Music] welcome to zero knowledge I'm your host Anna Rose in this podcast we will 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 this week I catch up with Yin Sandy from scroll we check back in on the CK evm L2 project and find out how it's developed over the last year we talk about the founding of a fully remote organization the technical evolution of the project the ecosystem and more

a quick disclosure I am through the ZK validator an investor in the project and it's been fun to learn about their progress now before we kick off I do want to direct you to the ZK jobs board there you'll find jobs from Top teams working in ZK so if you're looking for your next job opportunity be sure to check it out and if you're a team looking find great talent be sure to add your jobs to the jobs board as well I've added the link in the show notes now Tanya will share a little bit about this week's sponsor ever feel like developing zero knowledge proofs is a daunting task while the team at risk zero is here to remind you that it doesn't have to be that way they're out of the box tooling allows developers to access the magic of ZK proofs from

any chain without needing to learn custom languages or build custom ZK circuits Bonsai risk Zero's most anticipated product is a proving Marketplace that enables any protocol or application to leverage fast CK proofs in languages like rust go and C plus visit r0.link forward slash ZK podcast to learn more and sign up for the Bonsai wait list you can also find the link in our show notes and now here's our episode [Music] so today I'm here with two of the co-founders of scroll yay and Sandy welcome to the show thanks for having us Anna very excited to be here yes thanks for having us yeah you were on last year

in I think June 2022 we did an interview with yourself and your third co-founder Hai chin so that was like our introduction to the project on the show and yeah something that I'd like to do with this episode is definitely do a catch up before we do that though yeah why don't you quickly introduce yourself sure um hi everyone my name is yeah I'm the governor of scroll I work on ZK research Hardware acceleration for their own proof some algorithm side and some application side nice and Sandy this is the first time uh you're on the show so why don't you tell us a little bit about yourself what got you interested in this space and kind of where was your sort of Journey to scroll where did that start yeah no um thanks for having me

um so my my crypto Journey started a long time ago around 2013 and 14 that's when I first read the Bitcoin white paper and you know had that aha moment and just been kind of researching and dabbling and socializing in the space um I my first kind of full-time crypto job started in about 2017 um when I became a crypto investor full-time and the journey for Scrolls started when I was getting deeper and deeper into protocol level research and um and it became very clear to me at the time that ethereum was gaining traction both as a as a protocol and also on the

social layer there were just that the amount of talent and creativity surrounding the main protocol was astonishing and I I kind of came to this realization that this was the future of the crypto space and I wanted to see whether there are any kind of major open problems I could contribute to and that's when I met Yin Hai Chen we met online and we didn't actually meet in person for the entire first year of working together on scroll and yeah but you were not anons you knew each like you were yourselves I guess yes yes yes yes okay um well yeah couldn't have been an Anon because when we met um he had a paper pending publication on prover acceleration and that's how we

originally found him and and um and and haiten was working at Amazon at the time um you know as a system architect and he worked on a lot of the stuff that we're currently using for scroll hmm actually interesting the co-founding online how like where were you chatting how would you have like met and almost like vetted each other in just just online yeah I think this with with every kind of meeting there's always a little bit of serendipity involved how does anyone meet anyone in this day and age so it starts with with an online conversation and then there are some like offline connectivities and and personal relationships and um so

initially I was connected to year through um uh through through one of his kind of um one of our mutual friends and um you know started talking online for for a really long time and then decided to kind of um go for it you know what was the scroll project at the time like almost almost like very quickly but it was a research project and you know and and then just you know it just kind of continued to snowball from there already so I really liked that idea of like founding a company purely online where you don't I mean I guess it's like what year was this that you were um starting to look at this project it was during the peak of covid so it was about 2021 and I said it's the same

process for most of the people on the team um I think there were there are at least half the people on the team whom we none of the co-founders have met in person in the first you know and it's still very common for us to yeah yeah not meet someone in person until you know three or six months after they've been working with us um we're we're very decentralized as a result of this kind of type of origin and we have members from every continent apart from the Antarctica and um so spanning yes yes [Music] very keen on you know taking over that that um yeah piece of

ice um but yes um yeah so so essentially um I think that's part of our founding DNA and that's resulted to um you know scroll core contributors being um you know all over the world really cool and you were so you sort of mentioned that you were excited about ethereum this is like the project is kind of coming out of the ethereum ecosystem but like what was like had there already been proposals for zkevms at the time or things like scroll that you were playing on or you you know kind of being inspired by um I can talk about it from my perspective I think there's already been a few um l2s that are ongoing when we started um we were inspired by a lot of years

research and a lot of the research that were coming out of Academia at the time and um we thought there was something that we could offer both by means of Outsourcing proofs offline and also this way of organizing a completely open source kind of effort so maybe you can talk about his experience as well yeah yeah I think by the time when I started as Sandy mentioned there are several Layer Two efforts but I think none of them are really building a like evm by code level uh Vicky VM so some I think the most the closest one is building a language compatible but I think we are targeting a deeper level of compatibility which means we will suffer from a larger

problem overhead and like much more engineering work and we we kind of make sure that we can through Hardware accelerations through this crypto like Advanced crypto algorithms can make they efficiency become like order of magnitude like better than before so that's why we start this and with the deepest level of compatibility had things like Jordy and the Hermes eke evm proposal already come out when you guys started or what was it happening at the same time like I know that there was like the sort of I think I don't know if starkware proposed something but it was like I know CK's thinking proposed something but there was all these different types of zkvms but I do Wonder

was there anything like yours or was was your research kind of like really this new Direction yeah that's a great question so I think the very original story that we start with something our own version of like how we build a virtual machine maybe the KVM or like the virtual machine um and then we can make this really efficient through Hardware so that's our initial version and then like I talked with Barry White hat at the GM foundation and he's thinking thinking of exactly the same idea of building the KVM for ethium layer one and I think our idea overlap a lot for like why we think sporting a virtual machine is possible

and then so our like collaboration started very organically like with with Barry and with the CM foundation and and I think there are some nice idea for using lookup tables to handle the virtual machine memory part of the KVM I think the same idea also inspired Jordy to start the KVM so I think this kind of comes from yes yes same original Source but like kind of happening at the same time yes exactly cool yeah actually I should you know we will add a link to the previous episode what I wanted to actually do at least with the first part of this was do a little bit of a check-in on some of the decisions that were kind of being discussed back then

like I wasn't sure if those were like finalized or if they were like still in the works or if things had maybe changed so just a quick throwback a year ago when we talked about kind of what was the underlying ZK stuff it was like plonkish arithmetization and Halo 2. is that still what scroll is based on or has it brought in I don't know some like new techniques or other proving systems yeah I think think the text text deck is still like remain the same like we are still using plant cage for writing our third case and using qdg at the back end using Halo 2 to write our circuits but huge Improvement have been made on the completeness performance and the robustness so for example on the circuit

side I think when we talk we implement the majority of the circuits but there will still be some circuit left and by now we have already implemented all the op codes all the arrow cases and they have already been sent to the Auditors to to audit and on the plural side I think we make extremely fast GPU implementation to make proverb really fast and also we publish our paper so I think by the time we talk we only have one paper talking about how you use fpg and Asic to make poorer faster now we have published another paper talking about how you use GPU to make poorer faster and also we've done a bunch of liquid research reducing the memory consumption and also the recursive cost so basically I I think the day one we are running our the key via a full

version we require like one terabyte of CPU memory to run such a large gigantic circuit but now we reduce that to only like hundreds of gigabytes and we have to keep keep improving on that um and that is for for the two links I think we are using some derivative of Halo 2 which is called Halo 2C it's C referred to Community Edition so it's a derivative version from the cash Branch um and we because this is make we make many changes to this library and also we want to because it align with our philosophy to always develop with the community and sharing all these like proven stack and all the tools with the community so it's a community effort to own this proven stack and we are leading

some discussion to add more components into this Library I think a lot of projects actually we're using the same print stack like people are contributing to that even for some new like optimization from from fry it's totally different but people are still implementing like Goldilocks field frag gadgets to this to this community on the library um so yeah that's like a line with our philosophy a lot um yeah that's on the keysight cool I want to check the naming really quickly Halo 2 CE yes is that what you said what does that stand for Community Edition Edition okay and this is is this sort of like I mean is this the work that like PSE the what is it privacy and scaling

Explorations group from the EF and Xerox Park had been kind of like Champion or is there one track is it yours like did it also Branch out into other ones so originally we are using a fork of Halo 2 and the Upstream is in the like pse's uh GitHub Library uh and we we just changed the the polynomial commitment part from the inner product argument to kdg but then we want to make this more modular and shared by even more projects so there depends like you know a separate organization called CE and there will be more project contributing to this Library like Freedom XM and some other like you know projects are contributing

to this and we are leading some kind of more open discussions about how we can improve this Library so initially it starts with we want to build the KVM we want to use kdg so we use this fork and then we want to make this more modular and like more people can use that so that's how it's Branch out how do you sort of split your time in that regard like as a development like you're developing it like something you have a goal but at the same time there's like a lot of research and work that could potentially contribute to other projects like would you say you as a research team were like super focused just on the things that you need or do you think it's a little bit more fluid it sounds like you're also contributing just kind of more generally yeah I think our philosophy like trying

to support open source and more Community Driven and shared shared like tuning stack so we are happy to kind of either contribute to this like shared infrastructure and even Upstream to to the Halo 12 stream which is initially developed by the cache um so yeah we definitely like spend a lot of effort on like how we can uh like maximum like different projects like you know progress compatibility and also like because it also enhance our security right because if many people are like reusing the same proofing stack like there will be more people more eyes look into your code base and review and you also get a share like you know yeah so that's our philosophy that that's actually an interesting um incentive

there the idea that like you may actually just want to make something more General to be used by more people yes so that people try to break it early right I feel like there's been some chats actually in the ZK channels right now about like security of some of the systems that a lot of things are being built on and I do think if you can have more eyes on it obviously that's a really good way to like hopefully spot bugs early you talked a little bit about the GPU implementation this is for the verifier or the prover oh this is the for the for the approver right how you can make okay improve our Channel proof faster and when you talk about a GPU implementation Is this different from the rest of your Tech stack is this like some like lower level like how to better interact with a

GPU and I guess the second question is to this is like do most Roll-Ups in ZK evm teams need to do that because I don't think I've really heard about like the teams themselves working on the hardware so you can imagine that you know when you are running your algorithm initial on CPU for example it takes like one hour to generate the proof and you think it's it's too slow and you want to to you you utilize the parallelism inside the approving algorithm so you incorporate the GPU into your like CPU device and like you know it is just connected and then you will Outsource some very expensive computation to the GPU and gpus is very good for doing some parallelization work and it can do that

very well and the most computational habit part inside the Hoover Uh is paralyzable so it's very easy to clear that on GPU it can make you know become like 10 times faster so you are basically the high level idea that you move the most computational intensive part to GPU and the GPU can make that faster yeah that's what you're doing like on the software side you've made the decision what needs to be used by GPU and what exactly but are you actually dealing with like how the GPU process this is it as well yes yes so we also need to implement like how you kind of run this on GPU you need to design the kernel Implement using Cuda so yeah oh wow and yeah then I guess the second part of that question is is that

something that a lot of ZK VM teams need to be doing or is it because you had this background in Hardware that we talked about in the last episode that you were like oh we're gonna also do this hardware part yeah uh I think for GPU it can make all that they can become faster but I think currently currently from what I know mostly given teams are using gpus but some some that need uh because it depends on like your current algorithm uh if you highly rely on some kind of uh if a curve based operation which can be massively paralyzable and can be accelerated really fast on GPU but if you are relying on more like smaller field hash function then maybe you can run that on CPU efficiently but I think

overall if you have GPU it can always be faster in most cases okay so I guess most of the ZK VM teams if not building it themselves will be using some sort of Open Source yes version of this because it's kind of an inevitable thing do you feel like this is just part of the engineering challenge in general like this is where you can get some improvements so you will inevitably want to go for it yeah I think it's definitely a part of the engineer effort but also like One Direction we are working on this we should also align with our like decentralization philosophy that we want to reduce the requirement for uh like you know for example the memory requirement for for GPU and we are trying to lower that requirement for example like maybe miners because people believe that you will move from prove our work you

will have a bunch of gpus and the One Direction might be how we can adapt our algorithm and generalize this this organelle B10 to adapt to other cheaper gpus so they can still reuse their machines to run our proving algorithm like I've done a few episodes recently on Hardware uh kind of looking through the lens of the zprize efforts do you feel like do you expect sort of just like a standard for how gpus and zkps are going to be used or do you think that there's still a lot of room to like develop unique Solutions there yeah that's a great question so it depends on the algorithm I think eventually it will become a problem which involves software and Hardware called design so for example like you know recently people

are really crazy about Nova and you might need some special components for making NOAA faster so because they're proven algorithm I keep you know involving so it's very hard to say that you know oh this is your final you know yeah yeah version of of GPU but I think most algorithms are reuting uh almost the same like primitive like fft multi-exponentiation and some Vector level computation maybe some hash function so I think the large modules are fixed but if you really want good performance you need to turn like you know the workflow and for your for your proof system so I think that haven't been fixed it depends on your software

stack So eventually it will become I think a software and a hardware called design problem but then do you think it won't be GPU and you'll just like move towards fpga and then Asic like just since if if it becomes a little bit more you know yeah if I think even you can like figure out which things you need to focus on could you start to go in that direction yeah yeah definitely but I think in our current version we asked you more bullish on like using GPU for short term because firstly like people have more uh like general purpose gpus in health and they can run poor without buying like expensive RPG and Asic and also one interesting television from our conversation with different Hardware companies that uh single board fpga is very hard to beat GPU performance you

have to connect many ipg boards together to beat one GPU performance and async might be better so long term it might be Asic you know if if you are like kind of having one protocol become dominated protocol and they are easy for this protocol then the that might be a best combination but now like because the algorithm hasn't been fixed then the GPO is still General prepared people can buy that and so I think yeah short term we are still more yeah cool Sandy I want to sort of ask you a little bit about the kind of ecosystem and like scroll world since I mean you weren't on the last episode but like since maybe you started when you first founded the company was it just the three of you or was there

already a team and like yeah how has kind of the the world around scroll evolved um I would say day one there was just the three of us and um but the Snowball Effect happened very quickly um and I think the first turning point was when um you know I think as yeah mentioned that when when PSC team when Barry supported started supporting the project and we very much became a global Community effort overnight and having that you know very strong partnership meant that we're able to scale and gather resources in a way that wouldn't be possible otherwise and then I think the second turning point was when we

started having an MVP and when our vision started become clearer and the vision was to build a product that mirrors the developer experience on ethereum layer one as much as technically feasible and once that positioning is clear it started attracting more and more talents who are like-minded and who also sees the same vision I think I think that the wider vision is that we're in the first of nine innings of the you know Global you know if we think blockchain will be the value layer of the internet um then we're in the first of nine innings and if we're able to kind of

have the leg up of inheriting all the tooling and all the great things have been built on ethereum over the last few years then we're off to a we're off to a good start but it's still a start so I think once that vision is while articulated and you know shared within the wider ethereum researcher and developer Community we started to see a effect where people are applying to scroll and people are reaching out to us on Twitter and I think from early on we had this policy to hire the best talent globally so you're not restricted by where you live and you're not restricted by you know your identity and obviously we do kyc on anyone that joins us

full-time but essentially we started building this culture of like hiring straight from Twitter and hiring straight from Discord and and that was very energizing um for a lot of our community members who can see people who are actively contributed to open source research so this is kind of I think one thing I often explain to web to people how scroll has grown so quickly in head of head count is that you know the great companies from 10 years ago are built from a very kind of employee corporate structure but now we're very much a kind of you know we need to articulate our vision and set up in a way that anyone can see what we're building kind of week to week and anyone can catch up and contribute as they as they wish and if

someone does have value add to to the process then we'll bring them into the onto the core team and train them and equip equip them with whatever they need to like succeed in this endeavor so I think that's how we quickly scaled and we have a 40 plus engineering team now and I think it's it's very difficult to find ZK engineers and especially ones with prior experience and um and I think that's that's very much due to you know the you know one is the the division and the architecture that yeah has so articulate has so well articulated another thing goes to kind of high tensibility to you know run an agile process run a global development

engineering team and and making sure everyone is you know well supported to work on this really yeah how do you do that like especially if you're you're because yeah you were you're more I think of you a little bit more on the research side yes is Hai Chen more on the engineering side yes exactly our engineering team and I'm more on the research side because research if you're like getting kind of contributors in are you growing in a very like flat way or is it sort of like you're starting to create these structures yeah I'm just curious how you do it because like I guess how big were you last year I think last year when we talked it's around like

it's still like 30 or something researcher engineer but it's yeah you've grown by 10. okay so it's not too much too crazy yeah but that's only only technical but maybe just from that starting point to that yeah like from that starting point of just like a very small team to the 30 or to the 40. actually I think there's a lot of teams out there who could really like who potentially really want to know how you did that because I think there's sometimes um it's hard to maybe create those structures on the fly as you're going really quickly um we have lots of mini teams it's not like you know there's the zkevm team that that collaborates with the PSE very closely and then there's the infra team and then um the the hardware team is part of the infra team that talks very

closely to them and we try not to have teams that are across more than two time zones in order to make make sure everyone has a good work-life balance and um we're also starting to have some kind of regional um kind of city offices so we don't have any City whether more than three scrollers right now but we're starting we have three in New York and we have three in San Francisco so we're in the process of setting up like many kind of co-work offices and we're hoping you know at least a New York one has proven to be a a reasonable kind of hub for ZK researchers to stop by and hang out and we're hoping to have more of these types of kind of pop-up offices everywhere nice

um yay on that research front that you're kind of more in would you say like you have your inner team the folks that are on the scroll team but I feel like you must I mean you I've seen you actually collaborating with lots of different groups how does that all work and like yeah I'm kind of curious like how much of the new research are you letting in like is there a point where you have to kind of cap it and say like we can't do any more changes we can't add any more of the like cool new stuff or are you do you feel like it's still very very fluid yeah that's a great question I feel like there was two questions in there sorry there's like there's like what's it like working with other researchers outside and inside and then there's yeah how are you

incorporating new research yeah that makes sense yes so the way like we are we are like all pretty on the research team is that we Define some larger scopes for example like their ziki research working on like more efficient poor more efficient aggregation and uh yeah definitely looking into like new constructions like recently people get no appeal than the whole thinking about how we can move some Primitives to Nova and combine Integrity in our current system so that's on the research side on the protocol research side we are looking to for example decentralization of prover and sequencer the mechanism designed between and how we handle Mev and also like how we use multi-prover to improve the security and maybe

eventually when we like you know when you see a move to like tank sharding we will be bottom like by this execution we are looking to event position and more efficient client implementation so they are actually and also like some resource pricing for folio too so there are some large Scopes there we are always looking towards a problem we are going to solve in one years or two years so we're actually working on that but you know you also mentioned there is a product going on there is like you know you always need to bounce some stable version you need audit you need stable so we will have multiple versions and in this uh in this like recent version yeah stable branch which engineering Engineers are mostly working on and they are looking to more stable light and the

battle like you know better tested well audited as our version one and that is the next version we'll think about how we can like kind of merge those different research results into those kind of our current version and how they can integrate with our system and what's a kind of change a part we need to audit and how we kind of coordinate across different teams so it's more like you know even for ECM it has eips it has like small changes to make that works and in the next version so we will also have versioning and uh so it's like version one it will be very stable sound and complete the KVM but the next version might be like you know we make

some tricks we change the the pruning algorithm a little bit and then like make that even faster so yeah cool I want to talk about so this sort of speaks also to the sharing of research but like the competitive landscape last summer in I think at HCC there was like all the announcements of zkevms right it was like the week of ckvms and I remember that like at least the Scrolls he came in uh announcement I think you kind of put yourselves on the stage with that I think a lot of people didn't really know about what you were doing and I feel like this was kind of like an opener how has that changed tell me a little bit about you know what you're seeing on the competitive

landscape um I think I would probably challenge the framing that all l2s are competing in some fundamental sense I don't think this is a productive way to think about the space we were never thinking about the competition when we did the initial test net launch and we're not thinking about it now what has changed is that we're we're more and more confident with our test net results our our test net with proofs have been running for a few months and um and we can talk more about the roadmap and what we've learned since the Paris announcement um but you know taking a step back I think that the fundamental view is that there are still very few users in the space so there are like 2.5 billion monthly active users on Instagram for

example and that's just one single ad and there are five billion active internet users in the world so I think crypto as a whole probably has about like a million active users and and even that's being very generous with you know what we believe to be users in terms of definition so if we believe in a feature that has potentially a thousand X users then does it really make sense to think about like competition endlessly um I think we're all just kind of you know in the first of nine innings and and I think the real landscape is that um the the you know the space to be gained is so large and the design space for infrastructure to tackle is also

incredibly large like the feedback we've had are from a very very small subset of the internet population and whether that's real or true and whether that's like enough for like a real world testing that that's still like you know can be can be somewhat called into question so I think and and also another thing that's changing that's not related to the L2 space in particular but in the wider Internet space it's just that you know the definition of user and developer have been framed very differently so far in the crypto teapot in in the crypto bubble these two categories are framed very differently but in the real world you know every user in the world will eventually become a developer because of natural language

that can be used to produce functional you know programmatic instructions and so so the line between these two are blurring and the number of active developers you know like dabs and infras and so on is about 100K in the space and once again that that can easily double or 10x because of this particular change that's parallel to the crypto like universe and so you know we don't believe we're in competition with anyone I think the landscape is still evolving very rapidly and we care about our users and we're building and we're set up for longevity and for impact and um you know we actually pay very little attention to what the next L2 is doing but overall I think having more participants in the

space is very bullish for ethereum and it allows for more experimentation and that's good lesson for all of us and um and I think one thing that you know even if there are like hundreds of Roll-Ups you know when we go on the call again is that I think amongst all the Roll-Ups um I think legitimacy is the only thing that truly matters when you're thinking about trusting your life savings or trusting your assets or value to a network at the end of the day um I think the sense of like security and trust and legitimacy um although these are like very ambiguous terms but but what you trust a network to kind of put your life safe

into it it's a very nuanced and multi-faceted thing it has a lot to do with the development strategy and um you know the trust you can build and maintain and the systematic things you set up and and and I think it ties a lot into the core values like you know decentralization security and like well thought out design principles Etc so I think we're on this in the process of figuring out like what that is exactly I really like what you just said in the the sort of focus more on the collaborative part or growing the pie instead of trying to split the pie I still think in this space there is a company competitive nature some teams I think are already at least competing on

the marketing front do you see any benefit in that like it sounds like your strategy is to definitely focus on collaboration but like do you think it also brings a bit of attention that might actually be good um I think any attention to the L2 space is is you know net good for the for the world as a whole um I think the fact that we're seeing like you know centralized exchanges now issuing l2s rather than their own l1s is a move in the positive direction it's a sign of the wider crypto ecosystem adapting to the mindset that L2 is the way forward to is the way for scaling and so anything that moves people away from the alt one rotation trap I think

is a is a good step forward and I think any kind of marketing in that sense um is is positive that's interesting I just I have a question like I feel like the scroll project is born so deeply in within ethereum but have you as a team paid any attention to Cosmos or like the polka dot models like any of these other sort of like networks where they have the connectivity through different ways I mean I know the project is very ethereum is there any cross-pollination of like ideas between those ecosystems and ethereum um I would say I learned a lot from the polka dot ecosystem in the sense that I think there is a culture of putting developer first and I think that's also

part of the ethereum culture as well and and you know the things that made polkadot so successful we're also Keen to kind of learn from those lessons and and do the same for scroll just putting developer first and being steadfast and being a kind of friend on the developer to to found our journey as well um from from the from the polka dot ecosystem I think there's also a lot to learn um I think the lesson there is that certain projects have specific needs and um and and sometimes they they would like to share economics at the protocol level and I think these narratives are yet to be played out but one thing that keeps me building in the ethereum ecosystem is that if I if I have to spend a day figuring out how to bridge

stable coins into any new l1's then I'm gonna give up and I think that applies to both of these things um both of these ecosystems that that's my test on how long or whether I should kind of consider seriously like spending a significant amount of time or resources on on any ecosystem so so that ties into the second point I think is a it's a value that that we try to kind of prioritize as well which is just putting developer and and also user experience first and making it as accessible as possible so um you know hopefully that the process of bridging stable coins onto scroll will be you know a lot easier than than you know you would for

polka dot and interesting Cosmos I kind of want to ask a bit about how you envision the scroll roll up looking do you actually imagine it as a single instantiation or do you almost like could it be that there's like multiples of it four different uses I don't know if that's like a roadmap thing or if that's even something that's been thought of like currently are you just thinking about it's like a single version that like that's where you want all of the dapps to live okay I I can't talk a little bit more from the the technical perspective so I think we start with like you know solving the problem of ECM because if you have the contrast to it it's expensive so that's why we want to build a platform that can

solve this problem like you know all the transactions all the applications can deploy here and cheaper faster throughput and then you have the security so that's the starting point and I believe that a layer so imagine that in in five years like you know because you know layer one will be even more expensive and all the interactions with the ECM application will be moved to layer 2 and because layer 2 as a network has this strong Network effect for you know application have this dependency and how many users you have so I think there will be like very few layer tools who will be general purpose and lie young directly and then there might be some other chances to kind of

build their protocols on top of layer 2 either as at layer 3 or some more creative form we don't commit to any future because you know it's like even layer three the New Concept proposed by people we are not making like any any choice amount like you know we have to think future is like one layer two multiple layer 3s but we are thinking something like more creative maybe there are some other form like you know can also extend our layer too but we believe that as a layer 2 we should you know firstly in hide all the kind of not only from the technical perspective We inherited the security uh we are we are we are cheaper with a higher throughput but we also want to enhance this kind of legitimacy this branding of issue and because all the users were interact

through there too you want to kind of keep the spirit to be the same and then like more people can pre like build more creative stuff on top of scroll you know like cheaper Miner and but still like different teams had badly until future um but we are still like looking to what's a you know because also like again like many many Futures are like you know people are talking about the future but they are hand waving about the technical details we are thinking more in the long term that we propose something which is technically feasible and also it's good for building like the entire like you know it's almost like an entire country and then like people can build their own like Island cities on top of school so yeah but you don't

imagine deploying two Scrolls that was kind of what I'm wondering is like just generally like with once you build the stack you could right like you have all the software yeah but I guess there's no real use case for that right now yeah yeah people can still Fork our code base for for clicking VM and launch their music hero but again like it back to the question like when Sandy mentioned when they're the hundreds of raw up like including all the forks and legitimacy and the trust you build for a certain layers who will be you know the biggest issue like biggest reason why people choose you and then there will be a like massive Network effect why people will choose build on you and also because we are building a ziki VM which are general purpose uh the key or our

platform so we are not targeting a specific use cases so all the applications will we are deploying on us and also like especially specifically for some D5 like they need this compatibility so which means it's hard to kind of launch many app chains and having the same compatibility same UI and ux for users so that's why we believe that there will be some dominant layer tools with strong Network effect and then like you know even if other people are working they need to have their own feature maybe tailored for specific applications but we believe that's part of like you know scroll family instead of something in parallel got it and that that's actually sort of what I I'm trying to sort of Envision this is like a very general like roll up

future kind of vision but like do you sort of picture these like large generalized l2s being few and far between but then smaller l2s still attached potentially to the main chain the L1 that are specialized and then obviously on these larger l2s you could also have that L3 in certain cases but I wonder like is that sort of what you envision like it has have people talked about that I think all the people are invaded in like a similar future either like multiple Layer Two multiplayer three like build on top of each other but again like you know I think our point is that um we are still doing more technical assessment for how practical this is and what's the interability for example between layer 3s and what the benefits

you get from there and uh so we we think there might be even be some more creative format even like not layer three not layer 2 but maybe some more creative thing like in between to build a future I think no one can re if 100 that the future is look like this but there are like some reasoning behind like having layer 3s on top of layer 2 for example like imagine that if all the activities are happening on layer 3 people need to deposit and withdraw they also want to do this cheaply on their base chain right if you do this for example if you build some special application as a layer 2 and then like your deposit and with joy is also unless you have some way to have this kind of user want to leave put all their money

on your platform so unless you are very attractive that user won't do kind of deposit analy with job very frequently then like most applications will do this still and if you're updating layer 2 then this will be very expensive and if you are layer 2 then like you know it's it's you're saying if you build on L1 it would be very expensive right yes if you are building your building your layer 3 on on Layer Two then like your depositing with job will become much more like cheaper and also like you know because people have the reason to put their money on Layer Two um and then like you know but if if you know for your own application people are just you know interacting and playing with that for a while and then like they still need to like you know do this

interaction with the base chain very frequently which will become problematic yeah and also I think if you imagine the future to be all the applications on Layer Two then like you know it's beneficial for layer 3 to launch on layer 2 to kind of directly get the users get the liquidity down there too and build some featured applications yeah there are already been some experimental like layer 3s deploying on scroll and also that's what we we plan to explore like um adding more verification algorithm on supporting that on our layer 2 to support more creative experiments do you actually do you ever look into things like eigenlayer or sort of The Da levels like is that something because that is

sort of being built for the evm the the l2s right in a way yeah yeah yeah yeah in a way that technically like you know if you use use other da Solutions it can be cheaper in some sense but they are still oh by the way by da we mean data availability sort of should have said data availability okay yeah thanks for explanation okay so yeah so I think for for us for our own purpose because the we believe that as a layer 2 platform we want to inherit all the security properties of of ethereum so that's why we decided to post our data onto layer one to maintain the same security because whatever the the idea solution you are using there will always

be trade-off like you know for example like if you become a validator uh like and you will take across different platforms there might be some risk for kind of you know if one platform has some problem then it brings some systematic security issues so I think for us we just want to stick to this very traditional model to enhance all the security models and become like by definition a row up because there are some like interesting database also around what's your definition for row up and I think it a more popular opinion is that you post your data onto ECM and let the FM decide which is your canonic chain and so which that's a reason why Billy posting this data

option will always bring your like highly highly security and maybe for some other applications depending on scroll they can decide about their own Community either they want to have their own da solution or running their own sequencer so that's another reason why people want to launch their earlier so they want to have some kind of yeah like governance over this process as you describe all this though all I can think of is like the bridging between the different relatives and I think in when I first had when we were first talking about this I always imagined almost like a one-to-one like these little Bridges but I feel like there are newer Solutions now kind of emerging can you talk a little bit about that like how

you picture like do do you still imagine people actually using the L1 as the transaction layer or do you actually imagine the l2s kind of speaking more to each other yeah I I expect Leo to speak into each other but I think it's almost the same hard to build Bridges between different layer ones like even if you are sharing the same I think people have the impression that you know if you're building a bridge between two different layer ones um and uh that's one kind of bridge and if you are building bridges between like you know if you are sharing one layer one you build a multiple Layer Two then layer tooth bridging might be easier but in fact it's still very hard to kind of bridge between their twos I think so I think my prediction is that because all the transaction all the activities will

move to Layer Two so that's the reason why I think Bridges between layer 2 will be more frequent but it's still almost the same hard to build Bridges between different Layer Two because still even if you lie you there are some ideas like where because all your state route will be posted on the same layer one there might be some way that you can read data from State Route and prove that you know but it still you know takes uh either takes a longer time to generate proof or you don't have very great kind of atomic interpretability so I think it's still hard I think it's your like honest product like space and uh so people are still doing more experimentation there what about bridging that isn't token bridging but rather just like message

passing because I know there's like some really interesting things around that yeah I think my my own prediction for they say that because arbitrary messaging is even harder than just token bridging because you can like make your kind of token transfer because it's there's more because you can make your token transfer become part of your arbitrary message right like if you want to kind of yeah some token to another layer two it's kind of part of your message right so it's I think it's even harder especially if it's like a smart contract function call or whatever so yeah now that's that's interesting I want to talk a little bit about I kind of want to revisit the ecosystem topic and you know the connection now that scroll is a standalone project with a

team it was grown out of the PSA group or like it was you know connected to the ethereum foundation and stuff like that but what is the Rapport what's the relationship like does ethereum support anything in scroll and then I also kind of want to ask about uh Xerox Park if there's anything because I I know that like at least in terms of the libraries there's a lot of overlap are there any sort of like yeah clear collaborations between those three groups yourselves and PSE and Xerox Park yeah I think yeah because part of our philosophy is like developing the open source way and we are really great for all the contributions from this community um and there are several like open source collaborators we're hoping working with as you mentioned like they're the PSE team and we have been called building the key VM over the same

code base for nearly two years and we we derived into a like our layer to the GBM but we are also helping them to build a ECM equivalent the KVM for foreign so that's the the relationship is purely we are in the same Community pushing the same dream yeah and uh so that's that's the relationship and we start together and we we have been like long-term collaboration and I think for Xerox park I think we we hosted many workshops around our tooling so that's why you see a lot of overlaps and why people are reusing the same tooling yeah because we have done a lot of like workshops with um teaching like people how to write Halo 2 circuits since early last year and also we give lectures about like

proven costs and our they came in a circulation so it has some impacts for you know the candidates there like why they choose this library and I think education is definitely super important for like why people are using this tooling instead of because you just get more tutorials and more help like when you put stress this yeah I think yeah I think our relationship is more like you know we believe in the same goal we are value your line and we believe in this open source and so that's why we came together and sharing some educational resources and discuss and more recently there will be a lot of discussions about new printing systems like folding Supernova paranoia hypernova like you know it's you know we find something to

put together and then we are excited to see like how far it can go and uh we also want to support this ecosystem and this community grow as much as we can like how people are like using this and how it can carry into our system I personally just really love watching and seeing how these ideas are developed like how like some very compelling proven system or new research is introduced and then you see a tutorial or someone wants to teach about it and I think the Nova example is a great one because I mean it was almost a year ago that Justin Drake was like I want to do he wanted to do a whiteboard session about Nova and I didn't know what that was at all and but now you can see like at least in the sort of PSE Xerox Park and you guys like you see this sort of

like there's a lot of people who are really excited about it and now you see that also spreading too I know Arielle got Nova pilled recently and like there's like new kind of there's people who are getting introduced to these ideas um and then what's it's interesting to watch that happen and the material that gets created but it's also sort of interesting to see like why are there proving systems that also might be really compelling aren't quite getting the Mind share and like as there are more groups popping up because I don't know if you've noticed but like there's new ZK everything every time like every week there's like a new ZK newsletter a new ZK event and that's awesome and I do wonder if we're gonna see sort of these like areas of research these sort of like clusters almost of shared research

I think this is one and it's really fascinating and kind of amazing to see it in action that was my little Community take there yes yes I think part of our filthy is like you know we want to geographically extend a sticky Community to everywhere I think Sandia have more or to say about like how you rebuild that this kind of sticky Community around every corner of the world and how we spread this kind of information because through education through some other local events yeah no I was going to add to that and say I think you know any team essentially that is um you know aligned with their open source culture and focused on public education can join part of this development process is not definitely not exclusive to scroll PSE and Xerox

Park and and you can see a lot of amazing DK researchers um you know as a part of this group that are expanding um you know whether it's new folding schemes or or Nova and um trying to but but I think that the starting point is that there is a critical mass of projects and and researchers and projects and and Engineers who subscribe to this way of thinking and we're hoping we're one of the factors that makes this kind of philosophy more widespread amongst a wider ecosystem so the idea that you have to build everything yourself or or like hold it back from other people you have to keep it to yourself I think that sort of mindset is you know slightly outdated and we want

it to be like outdated for the ZK world as a whole tell me a little bit about the road map for scroll like I think we've done a really good job of catching us up to hear what we what's happened in the last year and I think yeah you did you teased sort of what was coming and stuff like that but like yeah tell us a little bit about the road map where where is scroll at you said there's test Nets is there a launch date are you going to be careful actually like announcing launch dates yeah tell me what's up yeah I can talk more on the technical roadmap and Sandy can talk a little bit more about our ecosystem building um so on the technical side I think currently our first priority is that we are we will launch management uh like everyone's talking about this okay yeah

soon yeah so the first priority is the complete and the sound the key VM with the robust infrastructure which means test of tests of time and like battle tested by all the community members for example we our current test net have been like there are over like 20 million transactions it's a very large number and over like 1.5 blocks and I think it run like formula addresses or something like it's like a lot of people are like actually playing with our test net so making that more robust providing feedback and uh and also we are we already send our our circuit to the auditor so that's our first part building something complete and then next step will be decentralized approver

um so the point of designer approval is it's actually different from the like traditional point of decentralization which is censorship resistant but building a decentral program Network actually had two biggest purpose one is that make your network more resilient because poor needs a larger cost to to run right like people need to buy especially like Hardware maybe and if our poor goes down then other people there will be a bunch of backup who can still generally prove for you and also it can incentivize people to build better and better Hardware like you know if you are incentivized to run approver you are incentivized to make that more efficient to save your own money right like you know it can make our Network they this kind of finalization time on layer one proof generation time becomes

smaller and the next step will be design sequencer so even if you know as a layer tool even if you are running a centralized sequencer you still have some way to avoid kind of censorship resistance for example you can enforce some transaction on layer one to enforce that execution layer too but it's still good to have a decent sequencer because there are some problem of real-time censorship like for example you send like you are going to be liquidating in one second and then I reject you and then I will include this channel in one hour which is still bad so that's why like you want to have this and also some regulation and also like depends on your philosophy for how you capture mu value and what this kind of value and mechanism how that flow to two different

roles I think that's the three biggest thing our roadmap um on the technical side yeah with a with a very quick side question here with a non-decentralized sequencer does that sequencer collect all the Mev pretty much is it sort of like an unfair advantage to that sequencer yeah it has this power to maybe not all Mev but like a lot of yeah it has power to kind of collect all the Mev feeds it depends on like you know like whether you believe this layer 2 will kind of operating how you define it badly or not but yeah currently like you know I don't think like any layered tools are purposely like leveraging this unique advantage to

run Meb to run some bad things but there are also good enemies like like you know Arbitrage to kind of make uh like you know balance your price and something like that but but I think eventually all the layer tools are committed to decentralized in some sense but it's just a very complicated design problem because you need to consider the yeah the value and like how that flow and there will be a lot more things happening uh like like not only designer program sequencer but a lot more things also happening like in our roadmap which have been parallel including some efficient client implementation like for example we will think about having multiple clients um and having EV ampization having that more efficient and also we want to

enhance our Securities through multi-prover um and also like you know because if I'm also involving we are part of this large ecosystem we need to think about like after they have for example in protocol PBS which is short for product like proposal Builder separation if they have this infrastructure for accounts abstraction uh which is 4337 and how we can engage and how we can maybe even overlap with this infrastructure so that's a large problem like how we can coordinate changes with ECM layer one because one important thing which differentiates from other layer ones is that layer 2 can actually drive the changes of each mm layer one because we have the same goal to skill so that's something that we are also saying

thinking um the talking with the ECM ecosystem and talking with the assault leaders and how we can collaborate in a more open way and generate you know like generalize our research post our research and having more open discussion and even maybe in a future form sounds standard for our layer twos we're really excited by how much interest we've gathered in the permissionless test net that we've been running for the last few months um as he has said we have more than four million unique wallet addresses on Gourley which I think it's a it's a meaningful number and it's very interesting and um you know that there's so much interest in a test net and and also it gives us confidence that we're

processing over half a million transactions on a daily basis it means our code base is it's been battle tested for about you know more than four months now with no you know critical failures and um there's still a ton of like optimization to be be done but I think this is a very good basis and it gives us the confidence to start talking to protocols and ecosystem partners and and gearing up for for a mainnet as he has said um so on one part where we're trying really hard um to onboard all of the tooling and infrastructure providers that are existing on ethereum so for them the integration process should be no more than a day there's no difference from

deploying on ethereum very one other than RPC so it's a very quick process and but there is a conversation to be had and we're very actively having those conversations on the other side we've we've recently launched a permissionless kind of page for ecosystem projects to register them themselves and to so that they can find each other and where any prep work is required to collaborate they can use that as a forum and within the first week we've had over a hundred projects kind of submit their their information through this site and um and we're in the process of thinking about how to organize it a little bit better um and to provide like you know collaboration between various protocols

and just bring a little bit of transparency and on day one I think users will be very interested in you know what are the things that they could play with what are the things they could interact with and we're trying to build a very good kind of streamlined user experience so that traffic can be directed to whatever kind of you know fundrise they want to play with so to speak what's the name of that website by the way ah so you can find it through um the I think the easiest way is to find a build with scroll handle on Twitter it's tagged on the main Twitter account and there's a form which you could kind of go into and then it's it would be I think it's linked to the main website

and and I think over the last few months we've been hosting a lot of kind of local you know hackathons and also local meetups and how houses yeah you guys were also at ZK hacklist yes we were on squirrel I think there was the ZK ZK there's a this is I guess an L2 yes privacy L3 um I think um using uh yeah using a lot of the the Aztec um tool stack so that's very exciting and um yeah so there's a so so so I would say um you know one thing that surprised me in this ecosystem building process is that there's a lot of organic ZK dapps and ZK implementations and I

think partly that's because of our our team's DNA and ZK research and so so that's kind of you know through osmosis inspired a lot of you know ZK Engineers to build things that are adjacent to scroll and also on the other side um we have a very active group of grassroot developers and um you know they're just hacking away and thinking of new things to be built and um there's a there's almost like one you know one or two games popping up every other week and I tried to stay on top of what's happening and playing with them but you know the there's a lot of interesting native things that are being built and preparing for for our for our mainnet launch and and also I think that there's

a team that's working really um really hard on kind of bringing on the existing more well-known protocols onto scroll the idea is that we want ethereum users who are familiar with certain brands and certain protocols to find things that they're familiar with and they already know how to use so there's a sense of familiarity and then they can also find you know new adventures like new protocols to you know play with and and new games to interact with do you see gaming as one of the key spaces that you want to sort of develop in terms of like just yeah do you want more games more yeah blockchain games on there is that a focus category in terms of product we're building a generalized eke ebm so we

don't have to choose so we support everything and the the pros is that that means you know we offer something that's interoperable and um something that can interact you know if someone were to build a game five project on scroll it's more interoperable with the D5 stack and so and and vice versa so I think that's the interesting thing so we are seeing you know games as a whole it's it's easier to build in a silo and it has less dependencies and um and I can see that being very popular amongst more isolated kind of devs or or Studios that are more organized around a single game type or processes

um so so I think that I can see a lot of games happening organically as a as a platform we're just focused on building exactly the same platform infrastructure experience as ethereum layer one yeah yeah a lot of games have been known to thrive in [Music] there got it cool maybe we could talk about Hardware just very quickly I think um in terms of ecosystem building I think we're quite unique in the sense that there are like more than 10 Hardware companies Building Solutions for scroll and um and I think that's like one one example of Open Source collaboration that's reaped or or energized the whole adjacent industry

um well I want to say a big thank you to both of you for coming back on well ta to coming back on the show Sandy for joining us on the show it was really great to get a chance to find out what's been happening in scroll this past year it's a pleasure to be here thanks for having us yes thanks nice to meet you again cool and I want to say a big thank you to the podcast team Henrik Jonas who's editing this episode uh Rachel and Tanya and to our listeners thanks for listening foreign