AMA with Vitalik about Ethereum Protocol Roadmap

[Music] my actually this event was suggested by oh I have idea can do about a technical is being recorded by yes okay so okay so I guess I great I I won't make fun of some of the other greaters too much that research this is research yes so uh so a small group they we focus on the no Ms um so we collected some question but I think it's uh question are also in your so um today we start with asking um so thanks and anyone want to share some inside recently rep a tweet about what you think about the crossing uh whatp cross and uh can you kind of explain what do you think each played in right yeah it's a good question so the way that like I think about the high level goal right is that like we basically want to try to get all of the benefits of like ethereum being like an ecosystem that is a unified ecosystem you know like it does it does charting like everything is everything is etherum have different like etherum but at the same time still having these player toos that are trying different things with security models execution environment and so on right and so the high level question is basically like what is what is they like the 020 like what are the few things that we need to standardized to like really make the highest impact in terms of like solving the problems that we have now which is basically like the ecosystem looking like 34 different watches and so the ones that I've been focusing on are so the the one at the top of that list right was usp37 data which is basically like address formats where the address includes know the chain or like and then like have the address after it and there's actually like some discussions that's still lot going on like there's people that have been very recently proposing like different versions of 5 3770 like there's things like know do you put the chain at the beginning or at the end there's that you know do use an offchain registry or to be just a regs if you want to have an address that's valid on multiple chains like to a standard for that so there's like some discussions for this right but might be yeah underlying an idea is basically that that and if you give someone an address like that by itself should contain the information that they need to actually pay you in a way where you actually get the money right and like that should be true for a person and for an application so 3770 is like just an address format and if that just gets adopted then that's a a really big part of part of the whole problem solved already right because and especially if you also include wallets being able to automatically what cross chain said said if like if you put in an address and you're 777 address that's on a chain where you don't have any assets here right and we get to the point where wallets can like actually just like understand that they're supposed to do a cross lers who transfer and just process that transfer automatically right then like that solves a lot of problems right so that's like part one the next part I think is uh so we talked about uh I I brought up 7683 which was like a standard for basically creating orders so like basically a standard for like doing a thing where if you're moving points from chain a to chain B then like on chain a you put your points into a smart contract like like the lock boox contract and then the rules of that contract say whoever submits a proof that they were the one that sent me like some other points on sh and like if that proof comes like that's okay then they can get the points right so it's like basically like order going going cross I think uh I've been like talking to some W people there have been people who proposed like some different uh alternative ideas right like one of the alterntive ideas was instead of restricting it to assets like maybe may be a call instead and so like you basically got like on one chain say you know I'm going to give a bounty to whoever proc was this call on some other chain like that's nice because that also lets you do things like you know only have gas on one chain and then still be able to send the transactions anywhere and so but then there's like the still the question of know what needs to be what do we actually need to standardize and like what actually needs to be left to like individual Wellness right and so like the way that I think about this right is that like what I don't want to see is want to see like 10,000 different applications having to like all do you like custom code bring in custom libraries that like interact with some kind of like centralized intered training right I think like there's probably more minimal ways to do where you just make a standardized I don't know what the API call that just says like I wants to do this cross chain thing and then it could be done through 763 it could be done it could be done some other way so I think there's still some not discussion there I think the thing that I wants to focus on right is basically like the top level gole right the top level goal is basically that users from the point of view of either a user or from the point of view of like adap like JavaScript code it should just look like you know like the concept of a cross chain operation should just block box right so that's the end of the the goal there then uh the third one I in was uh 3668 right and that's like the cross chain like like proof verification right so the idea there is basically that uh like we have uh right now we have L1 and know we know how to access we know how to prove things on L1 but let's say he wants to prove I know like some piece of stage on sub L2 and there's like a few different reasons of why I might want to do this right so one reason is if let's say you have an ens thing like let's say you wants to make your like you have your own application and your own application is like built on top of NS but you wants to have like S Sub domain and he wants to issue subdomains on your old La too right then people should be able to like actually be like do the ens resolving even if the don't already like locally R custom code that like is for that way right and so what 3668 does is it's like a standard where an L2 can define a read function and the read function just like takes as inlo the loal proof and then it verifies and that's like completely abstracted right so like if the weer two if they wants to use more they can if they want to use veral trees they can if he wants to use besid they can right like completely abstracted um and like basically from a users's point of view what you get is like you can resolve you know things like names on any Layer Two even if you don't already have code that supports that layer two right so the initial motivating use Cas was like this is a way to steal EMS but then the thing that we had realized right is that if you have that then you can also use this to make a l PL for any layer to right and so if you basically like if you have a l for one right if you have like something like cios which like syn verifies the chain then you can verify Merle branches and so you can verify anything on one but then what these one layers right like what what we don't want to see is like a world where layer one is like super decentralized me a l of sacrifices being layer one decentralized but then for Layer Two where like happens you know if you just have to trust our right and so the idea is that like real ially you know you still use rbcs but then rbcs can reply to a call and they can provide proofs and you can actually use the the onchain contract of the layer two or two mon bu those proofs right so we get like that trust withness and then the like to me the like the longer term goal is basically that like I think like I want to see a world where creating a new layer to wh permissionless and what I mean by that is like obviously make from now but I think there should should be some base level of functionality that you're ler to can just have automatically even if you don't talk to right even if all you do is you just like deploy a contract register ons deploy a read function like that already gives you some basic information that you need it to like able to read right to what to that right so like I think you know this like this is uh know one of the but basically know if you step back again right if you ask the question question if 2s did not exist and instead you know we have like let's say the 2017 Shing Vision right and like eum had you know, 124 shards and know let's say you know even different applications for on different shards like what what are the basic properties that you would still have right and like basically the pro what one is you you you still you should be able to write like you should be able to send transactions trustless way you should be able to R trustless you should be able to like P right so like those three things are basically trying to achieve like that and then after that I think I had like more complicated topics like key store wallets and longterm way or to stuff but I think those are like long question so very question you get any I remember like 5 years ago it was still treat ch right like oh you're not but now CH to SC part this do you think like the cross right I can I can use right ask because like um recently there's some Innovation invol fing like brand like do very stupid thing they CH used and unfor don't allow other like andate if you're L even if you're not the end right but since we have the stand you can say hey whatever right I mean I guess the way that I see this is like I get like I see there as being like different you know like conent so by standards that it makes sense to try to go like being in support right so like if you look at if you think about like addresses for example right like I think like I actually don't think there's much value in trying to like extend IRC 3770 so that it also covers Bitcoin and the reason is basically that like actually the Bitcoin World already has their own standard for I don't know like what address does it look like from what address is on different like bit deres chain and like the standard doesn't even intersect and uh know there like the cost of doing like that social coordination of like saying hey yeah you know we're going to get the entire ecosystem together like that's like it's much harder right than like just doing some doing something smaller uh for but then for the cross chain uh trading use case I think uh like what so the big difference in what you get for like layer 2 ler 2 versus what you get for like ler 2 offchain right it's like it's basically the trust Lo right because if you do something that's like fully offchain then you know you need to rely on some kind of Oracle to actually prove that know something happen on the on the destination chain right I think having that share formats is uh like something that I that I definitely think would would provide a lot of benefits um have so I think it definitely makes sense to have format have you know shared eventually get shareds that have formats they can be kind like now sh Astic as possible um but then you know on top of that I think we also need like an inner layer of Standards that's specific toer to so that you can actually get the benefits of like making sure that that like cross operations or trustless for example um like like there's longterm ideas that go even further than 7683 that we like basically get ideas where we can actually have like natively cross lay through tokens and for that to happen like you like that basically only works it's in layer twos that all commit to the same chain within episode right because like basically what the way would have to work is like ifer 2 would have to make merge more with like you know the same lay Balan right like this said an idea that I've heard from adjust and some like other more recently so you know if there's like standards like that and then another example is like if you think about that one of the ideas for 377 is that instead of using an offchain registry which is like it was totally fine back we had like 10 chains but it's not very socially scalable now like actually using an N Fring right and then like the level of ethereum connectedness you need to have is like you need to haveb right but you can have ANB and still be a side chain you can have ANS standing you like at the same time have a a totally independent ecosystem so it's uh yeah so like that's you know like again it's like a lower level of like connectedness right and that's something that like you know like Finance uh smart chain I mean I don't know they might they might not be willing to do but like you know chap for example right because they're like they they are you know in terms of they Vision they're like somewhat ethereum connected but they're also not like like they're not trying to be like that right so yeah so and then like another example kind of on the other side right is that there are lots of things that are like very unambigously Dr or but then like they don't even use EDF they don't even use TSA they're like you great so I think ideally standards would be like as modu as possible and know chains that are compatible with each other along you know one set of criteria and what be able to share the standards that makes sense for that and then chance be able to share as as much as possible as well uh I want uh I want to ask um see trees and stuff like what's your vision for solo state gting in like maybe three or five like everyone with a decent amount of run different Val on something yeah um okay so I think so there's a couple of related questions right I one question is like uh you know what what is a a level of like actual benefits of the network that Sol staking and small staking and realistically in and then another question is like what are specific things that we can do so start from the uh the status going through things for people so I think for the first question so this is a a recent like belief that I had right but like be don't I'll throw the idea out there I think we should increase the Forum threshold like the percentage in the off to finalize of what from 23 to 80% and I'll explain the reason why I make this uh basically because uh so first of all if we like realistically the cost to L this of that happening is like actually not very large right like I think the ethereum network like we only had one outage that goes below 23 and I believe we actually might not have even had any other that are like in% right or if we did like and and so the cost to we is is not high but then the benefit to safety right is basically if you imagine like a bad situation happening on ethereum which like two cases one of them is like software bug like y has a problem failure and then the second scenario is like actual 51% effect then the the less bad thing that can happen is finalization stops see have it can have speed or you actually have two Comm Forks right the reason why that's less bad is because we have two forks but because it's non finalizing like you're giving users the information that like hey you know these are right now on St change and people who need like a higher level of safety like they would already have written code that leads for finality right and then and like that actually meanwhile if you finalize something wrong like then that's like actually much more part com right and so like I think having a a wider space where we outcome is that finality stops for a period of time is like actually safer and so the reason why this ties into speak is that I think that getting like getting solo stakers to 51% is like right now it feels like an extremely farly W and like I definitely don't want you know like inch the get etherum like taking decentralization narrative on a vision that we can get to 51% Sol stakers if like right now we have a a credible road map for for get to therefore we can do is let's say we set the to 80% and then we can get to solo stakers and small in smaller institutional stakers so let's say you got bu Tok sing with their treasury um up to 21% right and then if we get that like if we can get a highly independent group of people that's like at least quum right then like in a world 10 years from now like where the ethereum is not run by like these like idealistic people that all believe in sens of resistance of the values but we you know do have big finance companies involved and that know we can't trust the social layer gra like if you at least have a very independent set of people being horri that like that produces the chance that ATT right so that's like the goal that I think is uh like both like like it feels to me like that's the goal that kind of strikes the balance of being you like both like actually very meaningful if we get to it but also actually achievable right so that's one part but then the second part is like how do we try to be more friendly to solo staking and small scale staking um and uh I uh like I've asked the question in many situations basically get like I think the boring question to ask is that know why don't why do you think more people build a mistake is that like puts you into the like abstract far mode right question ask is why don't use right and the question that I get is the answer that I us get is that basic like the number one thing is actually the 32 we minimum right like 32 we minimum is gets more hands than everything else com right and then number two is uh Tech technal difficulty and then number three is like all of this fancier stuff of like oh you know he staking when it with My5 or whatever whatever right and so priority number one I think is like so I actually think there are some really interesting strategies to reduce the staking minimum right and this has to do with orbit ssf this is one of these new staking ideas that got published on the research about a month ago basically the idea is like it's a uh it's like a random sample Community scheme right but it's a random sample Community scheme that's designed in a way that also has they get the properties that you need for like for a chain that has ecomic penality and like if this works out right then potentially know could way down to like making the staking minimum be the staking minimum be one without technology right so that's uh so like this is one of those things that like like I would actually want to see happen right and this is like actually one of the topics within the IND research group that's happening right now then the second topic then is you get speaking hardware issues right and like I do think that historically like we have a tendency to focus on the hardware side too much and say like oh you know if we have stainless at you speak without a hard drive and that's amazing but like actually the gain's not needing a hard drive or like not very high if the the biger limitation is like 100,000 Del right and so but I think you know our top priority is finding ways to like knock thatum down and then second top priority is also loosening the uh the hardware requirement somewhat um and then you know it's anyone to like find ways to like make make the speaking experience like easier even just on a p Le level right so like Bic like things like that not are good right there's a lot of these a lot of these other things that are good but like I think there's also like being like social part of the question of like how do you actually get like encourage and motivate people to like actually get like put their point into a staking box and realistically you know do the effort of like be sing every every time like the notification alarm be like something happened with the uh with the internet connection UPG orever uh so yes question about the Privacy Shield transaction landscape we've seen a few swings at that before we saw Network and sgx with B limits and we've seen some VM try to build the ground up for D Primitives and pl's going to wipe out a lot of the B based paing friendly stuff so that going me best I feel like the Privacy space has kind of turned into at least like two or three different spaces right because uh know if you look at a lot a lot of the stuff around like especially privacy that starts going Beyond as and starts getting into like multiparty compensation trusted execution environments fs and all that stuff a lot of the time their goal is not you know traditional privacy rules which are like hey get you know like I get you know like I Russia and I want to do need to opposition political parties to not go to 15 years right their goals are actually more on the side of like preserving Market integrity and like basically you get like having privacy of people's traing strategies protecting people from getting front run and things like that right and like I think there's value in like recognizing that are different spaces and I think you know they defitely benefit from some shared technology though even that's kind of starting to fray somewhat so I think for like the Privacy payment side right my like my own approach philosophical approach to this right now right like I think there's this there's this big transition that like we as an ethereum new system need to be from like the iCal and idea driven to like pragmatic and like how do we make this a realized benefit for users right and like you know my first answer on alter was basically saying you know what are some like very simple D ways to that for scaling and for privacy I think it's kind of similar right I think like I mean the question right now is like why like why aren't more people VI than existing privacy Solutions right and I think like there are answers to this but we have to like actually go through the answers and like actually take this very you know usage focused approach you know like Start From Start From basic things start from just you know fairly buy Modern standards Tech and unsexy things like just having magic math poal proofs for payments and then you once we have that you know start start going one level from there do you think do you think with the really basic Solutions we're going to run into a lot of architectures that's start look like and we're going to run into policy issues that way um I mean okay so I actually think that see this is the thing right the policy issues actually we come harder to resolve the more you get it to prob get stuff right so I can give one example right so if you think about Toro type pools right like the the current strategy for making them uh you know right is basically this privacy pools approach where if North Korea you know comes in cles you have some deposit delay and then you have some mechanism the flags to deposit and then everyone except the North Korean it's like makes a proof that says I'm any one of the depositors but I not North Korean and North Kore so we get any privacy right but then if you ask a question like let's move Beyond an in andout system and let's like actually have like a full in not sure where like everything's prod right and then the problem is like okay let's say for example you make a payment to me and you prove that you're not a North right and let's say I then have to take those points and you know I may have to make a payments to like CH go over there right but then let's say my own standards for what I want to prove are different right let's say for example like I also wants to prove I'm not you know like from the government of like a for example right and then the problem is that because you only need the approv that you're not from North Korea I have no way to prove that my points are not right and so you know we so like there there's this kind of composability thing where if you if you actually try to ask the question of like how do you make this design tolerate people having get like diversity of different perspectives of like which act is to associated with then you know it starts to become Bunch part right um and so like basically like these kinds of problems actually multiply more once you start trying like saying oh you know we don't just need pains we also need private contracts we also need full privacy preser defy and so on and so forth right and so like that's actually my like motivation for saying like let's start with application specific things and like the two things are basically one is like in and out payments and then two or these like very narrow you know not try to like anonymize participation which is trying to Ure marketing seg Ty architecture so like that you know get that to a Stam and then go from there first yes um so I have a question regarding ethereum road map as relates to um the search track so I see that you have listed a goal of like supporting up to 100,000 TPS um so we have Circle we work a lot with Enterprises that have like hundreds of millions of users yes um so they care a lot about TPS but at the same time do care about the gas so what we learn practically that even if certain chains CL that they have high TPS when you reach at before you even reach close at TPS that gas price surch it becomes impractical for any business to consider pay the gas that um so I'm curious that as you think about the road map for the Sur track and supporting that htps how are you thinking about balancing kind of like also lowering the gas cost and when do you think it would be um like there would be a date where we could imagine Enterprises have like hundreds of millions of users could use EDF chains on to support their frequent transactions yeah good question so I think um I think the answer has two parts because I think like the steal ability limitations that people have today have two parts right and so one part of that limitation is that we have um just like the evm itself right because ultimately like most layer twos they are poies in ethereum and if you like specific software orations of ethereum which were designed to run the ethereum L1 and so they never had a need to be able to process more than 15 transactions within a few 200 milliseconds and so like each individual to like has a hard time going above you knowas right and so I think the solution to that is like basically software engineering and like two kinds of software engineering right so one is like the hyper scale inm effort right so there's the effort that the Paradigm team is doing the effort that the Mega E people are doing there's other hper scale in projects right so that's like one type of uh solution and if they did the math right like there's no theoretical reason why you can't get like a single EP thread up to like something in 100,000 TPS range like it's basically a a software engineering test right and then the second part uh and then the other uh part of the solution is that there are I think there are nonm architectures that Mak sense for a lot of these cases right like especially for you know quote Enterprise blockchain which is like I know you know I concern that a lot of people hate because like it just got to we Tri and failed back in 2016 but I think you got try and fail back in 2016 because we is evil like the WR technology ST which is permission blockchain right the permission blockchains have the benefits of decent of sorry they have all the cost of decentralization because you have to worry about forums and tolerance and Ste eting but they have none of the benefits of UD of decentralization because like you're still L up for priv network not connected to it that and like it's not just Mone if like consum Swift as well like people still have the same concerns right and so like the right approach to do Enterprise watching I think is Chains It's M right like basically like you have your server you keep your server and then in parallel to your server you have like a piece of sidebar code that basically just keeps making pass of your database and proofs of your execution and it just keeps publishing proofs onto theum that basically say whatever is happening in the database whatever is happening here is valid right and like I think that approach actually like the one weird term use case that we that I really want to see is actually got group of solvency for centralized probability changes right because uh like if you think about it group of solvy per calized group changes is a super important blockchain right because you know you have points of need to constantly Pro to their users that they're you know that they're not being inent right so but if you're to do that then like your software development path is not start from have they griven your software development path is like start from your existing deployment and right a side car that like proofs whatever you're already doing go from right so like those are the proof paths so that's like one part of the problem right which is like hypers stealing execution the other part of the problem is like hyro stealing like bity basically right and like one part of this is obviously I know data availability and the current road map for data availability like the planet that try to go up to 16 megabytes slot which is 1.33 megabytes second and then and then a reasonably optimal roll up with have 23 B a transaction and so you know you divide 13 Million by about 23 and like get somewhere around you know 60k TPS or something like that right but that's for RS right um for and then the if you want to go even higher than like these like CL my thing is like still a very underrated technology right so like there's you know in Max marus Japanese them I is like one group that I really got respect they actually have like a hybrid R up plas arure where they put five bites on chain bir transaction and it turns out that that's enough to give them privacy and protection against like bad operators in way like in a way that plasma does not have right but then you know then one point you're basically dividing 1.33 Million by just you know like five bites and then you have I know 260 g DPS and then there's like other versions of plasa that can a little bit higher and then like if you're trying to you know do like a watch Chain video game thing then like you don't need like that more security you just do spr so I guess know like that's how I see just quick follow you mentioned that side architecture is there specific viip or like specific document that addresses that type of design I I I think in this case like it doesn't even need to be a orc right because it doesn't need to be standardized maybe the one piece that does need to be standardized it's like if you have these uni blockchains and they're getting users you like work groups of their data then like you want a standard way to that right and I guess that like grp 668 is basically already good enough to try to do that but otherwise it's like very much a you know like go off and build your own thing kind of thing which which I think isal right because I think know as an ecosystem we have some fairly know like limited like ordination resources but I think we want to like go the stuff that we where we AG is super valuable and like where we you know wants to UniFi and not have like 15 meting standards that like actually get pick a few directions and that nor job on them thanks going back to Sol state so uh how this in the future I want to keep sing and build own box yes uh how do you see band with requirements changing and it's realistic home not like okay so the thinking back in the 2021 era right was basically that like we all you know like surrender to this centralized block building decentralized inclusion centralized validation vision and uh like basically if you want to be a Lu builder then like you have to have know 160 you have to have the ability to just like push out 60 megabytes within a couple of seconds and if you're not that kind of act you become a elevator and then you still contribute to inclusion list and so you're you know your heing much sens resistance but noning so that was a thinking that I think since then you know there's been this like strong push from the community like basically saying like hey know we actually want like stronger assurances and so we have we have incl L right and then I think recently there's been this like near Universal consensus on like multi inclusion list providers right so instead of one inclusion list provider of course L you have like maybe eight or 16 of them but then the one thing that's like still kind of like in research land is making a really convincing design both for that and also what building side that uh doesn't require you actually have all of the data right for specifically for BLS right because for regular transaction when it's Cent problems and for blobs the basic uh you know way that all of these uh mechanisms have to work right is essentially that blogs get rebroadcasted and then you as a inclusion with Creer or as a builder like you can basically sample BL a sample and then if the sampling passes then you trust it Orly right like the way that the got the blob architecture is sort of intended to work from a software architecture standpoint right is a b basically you know you do this like Sly procedure and that lets you like you know put check put this check mark on one pieces of data and then from that point forward you just like treat it as you're working with a much smaller number of these pieces of data that out to less than and so on the inclusion list side like basically sample a sample passes you would add and then on the final buing side you would like do the same thing right and like the big subtlty of this right is that Das is only secure if you can actually guarantee that like enough people are data s playing at the same time right so that's like the one team that we have to really work through but like that's the the style that I think all the approaches think last question okay yeah the question I have in mind is actually pretty a little bit abstract you know way because I r post recently about in shining like what what features should we try to do at level and what should be you know what should be just like that that the community or any other people inside for us yeah and an interesting trend is like I I think what I saw is like some will came up like came up they will put some new on the table right in s for about years then somehow it shift the whole some everybody want to join and then it becomes some become something you want to solve because now you feel like um you know the rers or the the the flesh BS they dominant like they they can be sensored similarity to Li as well I'm just like wondering like would it actually be is it is it possible in the future that maybe one day we only have all all the users are um VAR that we have five l or something and then we start worry about these fire five layering right yeah like how how you approach this terms like what can what when do we decide what need to be in to the per right yeah I mean I think uh like I think realistically right there there uh probably like one philosophical difference with between ethereum and Bitcoin is like ethereum doesn't like pretends to be a fully a political organism right I mean big one I think he is still political but kind of pretends to be that like have like there's no such thing as like course developers of and like submit PRS to reach howy that other electoral sh works right but and you know this and and I think in system right like if there is one part of the stack that's like really not working for a critical mass of the community and people are starting to get like existential fears this is starting to violate like really decent properties that are why we came into the whole Space the is going to be get and it's a strong push to try to do something that way to try to resolve that right so I don't like I don't think that thing 100% go away I think the things that we can do today are one is like being proactive on the research right and like basically trying to like think of as we need problems as we can that kind of time and the uh then the other is uh to try to be proactive on like setting stand like setting good standards and expectations for what weer twos can do while they're still small and while they're still not relatively not aligned and while there aren't much more powerful forces that are like brething right and so I think in theer 2 case u i mean I also know you know this is a big part of the reason why not Justin is a big fan of Bas PS right because like jus who wants to just bu ethereum y Shar sequencing that includes as many l as possible and then if just divisions have SE then like basically any Layer Two which is automatically inherit the sensorship resistance of lay one for example right um so that's like one you know example of the other approach is basically that like the more modest thing right is basically the sign byass channels right we basically the layer two is one that in the layer two contracts they would specify an alternative way for people to send in cashes and transactions and then those just have to be included when the next time that someone publishes right and uh the so from what I've seen like mean there are rups that already have these bypass CHS alive when rollups are not willing to improve them the reason why they're not willing them is a lot of the time they're like actually pretty lame reasons and what I mean by that is in the case of ZK rollups for example right a lot of like sometimes the reason is basically that like they like there's some operations within the L2 that are like very difficult to prove but that are not expensive in terms of guess and so like you could have someone send a transaction through a blest channel that just ends up being like too big for them right and like the reason why I say this is lame is because there's like a proper way to solve this problem and the proper way to solve this problem is that you should basically way twos need to adjust their gas schedule so that things that are expensive for them actually are expensive in terms of gas right um and like I think there is coordination work that we can do to like actually solve that problem right and this is part of why you know I published in EIP to readjust some gas cost so there's discussions to try to make that Mak that to RS there's a lot of bench markting happening right so to the like I think there's a lot of cases where the reasons why l2s are lot better at this right now is not because they wants to not sensor anyone but it's just because there's like these weird other technical things that just are annoying for them and if we could just like come in and like help and basically yeah either make it easier for people to have different gas or provide standardized ways to do certain things whatever else then like that's valuable stuff that we just want to do like right now right because like right now I think no one is really no one just treating elers as being you know the type of actor you're supposed to that you're supposed to try to breathe down their neck right like El is actually again right now they're still relatively free to do thing and of course as time goes on and ases become more and more of a big deal like you know that's that level of optionality that they have is going to go down right and we're going to have to answer more to both political pressure and to their own to holders and so the you know like I think there's a lot of value in just uh like setting and pushing those uh and and reinforcing good expectations getting not getting good uh standards adopted uh very quickly right now to reduce the chance that know there's another one of these big STS in VAR two or 3 years in the future how do I see any words more centralized for now the yeah right yeah and so there's uh a couple of solutions to this right I mean one of one of them is uh and there's obviously this got like multiple proposer and CP approach that some people are like ma resting and like those people are trying to push and that approach basically said like it basically tries to make public vatory right it basically tries to make it so that like it's not even valid it's includ a transaction which just not the public process and like G in through you like pror biding and there's like a lot of people who are skeptical of that approach because there's just like two ways to get different different kinds of side channels and then there's per just the various positions in the middle so that's like what type of answer right and then the other type of answer of course is to basically say know it's like right at the application Level try to like find ways to make order order less relevant within a single block and I think there's like interesting ideas here that that have been kind of under explored and uh one of the uh things that that we can actually try to build on top is actually ERC 437 aggregation right because so if you think about the way 4337 aggregation works right basically get the idea is that you have like in different transactions they have in different signatures but you don't want to verify in signatures end times you have like some procedure for combining the signatures into one and so you verify only once Des right and uh what you could do what what this also is right is that you could could basically use the same mechanism except instead of using it for signatures you basically use it to essentially process all transactions that have to do with some particular application in a BL in order inde dependent right and so if you if you do that then like that also like reduces a lot of authority from you like what whoever the actors are that are actually that control right so like that's another example of solution and then obviously the third approach is to basically say that like you know it's said like basically status quo and just saying yet the you know the losses are not that bad and that like basically focus on trying to maximize sensorship right so like that's how I three of the pass work being but I think there could be more I think this is one of those space where I think people would like actually really welcome a lot more approach being propos and discuss right okay so I don't mean signature aggravation I mean like the technology that that supports signature aggrevation inside of us47 right because like so the way that it works but is you have like these different user operations and then if you have if they use signature aggregation then like what happens is that there's in that like basically there's like a validation step that's like P production but then there's also a validation Step at the end right the validation Step at the end is meant to check some kind of ACC signature right like this is how you BLS wallet works and like all these things right and so the idea is like it's What If instead of having the piece of the end be something that verifies signatures you have the piece of the end be something that just executes the entire Market L right and so basically what you have then is like you have order independ like execution at least within one type of application for at least all transaction yeah I okay soor signes the aggregation is to right like all the participants they need to talk to each other right and so I think for that reason like I generally consider it like it's like okay if you if for you kind of you like a multi type of use case but for saving gas between transactions that were made by different users it's not really viable right because like different users what they have different WS they not and talk to each other real time yeah let's look at these um there's some discussions about the side effects of introducing l space so higher Network man with I think uh yeah so okay so the shortterm solution to that is uh I mean I think there's like different so there's like a shortterm medium term and long term right so like very short term is that the way that LS are distributed even right now is like very suboptimal right and like there's a lot of just like very pure peer-to optim a that we right and like one nice thing that we have that we can build on top of is that logs already have irure poting for free right and so you can do things like if you have 32 peers you can like take the entire 2x blob and you can just broadcast one over 32 of the block so like all the different appears and then you only have two x overhead and you still have a lot of redundancy and then as soon as you get 16 pieces youu with a lot of like there's a lot of these like more Advanced strategies that are like actually not that many lines of code because we already have all the infrastructure for ra project set up because that's that you that's a BL for right so like there's stuff like that then pias right so pias basically lets you verify blogs without to download the whole thing and that's the plan for how we scale from the current approach which is targeting about 34 kiloby to the targeting like 2 megab right basically if we know we actually we request sampling download like give me5 or 10% of a BL then and then after that I like there going to full do and that then the question I think like splits in half between verification and like block production and then verification is the easy part because that's what data s what does and then for BL prodution that's the the question that I answer before now how do we need have more detalized talk about this account account based model versus uh sui networks uh objects okay to you too how terministic sequencing and or agnostic applications work together to reduce that in other situations where both are needed right um and then what happens multip Builders transactions but some Builders try to sensor and what you you know these are kind of three questions what they're right basically like would it be better to fix any problems at the applicational level instead of changing the protocol I think this is an interesting one right I think that like my answer is definitely you know in many in many contexts yes right like I think if you have an application that like leaks any Revenue then like that's not just a centralization externality on the network like that's cost that you are paying as an application development right and so it's you interest to try to find ways to reduce the the exposure and so I think I mean I've like come up with some like designs for like versions of the Miss SW that are more sish resistant I think Dan Robinson actually we we up of those from like four years ago recently there's a know things like C there's like order independent execution designs which I yet talked about I think like I'm super in favor of exploring on all of that stuff another really important space is liquidation right there money and other sorts of that that there's ways to try to make those be like less less dependent on you know being the first actor uh so I think it's like good to do those things but at the same time there's like limits to how much we can do those things um because that it's like at the end of the day if you have an onchain order book and at the same time offchain exchanges and offchain price this govern exists then like there's always an orage R opportunity with someone who has the ability to like add things into a blood like after everyone else right and so I think know accept these application layer techniques like they can reduce EV but they can't like fully solve the problem right I think like that's the reason why there's a desire to try to be the the protocol more robust to that right and then on top of that there's just people who are me with right it's like no sane protocol designer in the world would have proposed that like monkey sale that happens back in 2022 that just like spiked gas PES all the way up to like I think literally over 9,000 it just burned $180 million of easel in a day like do you guys remember that right but like no s protocol designer would ever recogniz that and yet it happen right and like theum needs to be robust to like people doing crazy things because it's a permissionless blockchain and so you know we get I think the it's like the application like side minimization is good but like the protocol side minimization I think is like also necessary um oh well then wait the questions black all replaced um how far does f go with practical applications I mean this is um so I feel like f gives uh like the kinds of applications that it targets ends up being like very different from the the things that CD stars do right and like sometimes they're separate sometimes they're complimentary which I think is uh really interesting right uh basically because what happens is uh you um like C St based applications they have this limitation that the only kind of private State they can handle is private state where the owner of the object has the private information there's like always one person who has the private information and that's the same person who actually has the authority to like action said transaction using that thing right with FH generally that one of the challenges is like you still like either it's like a two- party compute model where you're trying to solve a very kind of problem or you're basically using it as like a type of MPC and then you have an M trust assumption right and like M trust assumptions are risky because like the big difference between M trust assumptions with using like threshold anything and like 51% attacks is that 51% attacks are extremely detchable right like if someone does a 51% attack on ethereum one it's like obvious that % attack happens and that just HS everyone two it's obvious who participated and so you can be slashed or someone can start a minority soft for and you get inactivity right it's like very risky and very expensive but if let's say yeah you know we have a system where everyone's coins are inside of you know like some kind of you know like M MPC thing then someone could like go to 51% of the participants and they could get them to say hey I'm like install this like extra extra thing and then if they do that then it would be undetectable like which of the participants participated and it could even be undetectable that an attack even happened right like and so like that's something scarier like it's scarier especially if you shift your mental model from like the mfn being you know like friendly can be happy aligned ethereum people to be like mainstream major corporations right it's like you know if you imagine like apple Google and you like Microsoft like know doing a three of five being part of a three of five like you know if the NSA really wanted to like yeah they could way together like separately back to three them right and so that's like uh so that's like the reason why you don't want to lean on the end of that assumption too much right like I think there's ways to make it stronger you can like part on top and so on but like this is like to me you know the gold standard is like not having like that kind of assumption at all right but then at the same time there are applications where it's unavoidable right basically because uh what you so like there's applications that just inherently involve doing you know two what combined publication that combines multiple people's private data where you actually have to like combine the right like you can't you know decompose it into separate in dark pieces and so you have to do some kind of combined of like black compute that's as close to a black BL as possible one example of this isoing right like if you want to take something like Macy and then you want to try to make the coercion resistance based on a TR assumption that's like as minimal as possible then the best thing you can do is basically put like the ma decryption inside of NE aeve right that's like the best thing you can do right and then all of your trust all of the properties other than course resistance they're unconditional right because you have encryption we still have Z starts you still have blockchain sensitive resistance but like just the the co resistance piece becomes dependent on the seh Assumption like that's kind of avoidable right or another example is like some of these Financial Market Integrity use cases the like a combinatorial auction right and in that case it's like well if 51% actually start polluting then you know is it actually that bad um like it's bad from Market Integrity perspective but like at least um you know people people in Russia that are like trying to join into Political opposition are not going to go to jail right so you know it's like a a somewhat lower level of trust required and so I'm like more okay with that event there um if we're looking at the problem of censorship resistance that's our goal and looking at this new L2 landscape yeah and we think that we can do a cheaper and more data at similar level to some resistance how would we approach that problem you get published for many different LS and many week networks for mon a strong yeah so right I mean I feel like uh publishing to different uh l2s is like probably not like that that's definitely not the ideal architecture right because the ideal architecture I think is that El tools that should just all individually like have some kind of like strong multi proposing and like each of them individually should be get censorship resistant enough to satisfy our standards I think in the world of today where we have like lots of centralized sequencing it's like not great then if I was designing likeing application and wanted to like really crank sistance in the next and like yeah I would have the you know the ZK part like walk over as l2s as possible so if you want to you make your both any single one of the one of the l2s and like that actually does like you're basically using like multiple centralized L2 essentially as sequencers of sort of like a meta constructing and like that's actually a decent approach so I think yeah like if you're trying to be pragmatic and like give people guarantees in the short term absolutely I think in the long term we do what what like wants to make each two individual need yes off satisf property okay more pleas um future scalability Solutions particularly with the entation of charting and rollups how Envision the integration of these Technologies and improve the overall performance experience I mean I feel like we already kind of know about rollups I feel like we've kind of talked about this topic of like V rollups have the benefits of the you know like the 2017 Charing mission in terms of unifying ux I think the other thing that's important to also talk about is uh the you know there are scaling solution fund and rollups right so like got mentioned plasma I think is a very underrated category and I think like L2 are not just the scaling Strat solution they're also in that you know in execution experimentation solution and so like it's good that we have fuel it's good that we have azte it's you know good that we have c s white ma that whole spectum of of systems so like I think the goal of the all2 EOS system right is basically like strike this balance of like enabling a lot of this like right now you like like it's experimentation and then five years of the future it's like specialization of what makes sense for the use case but then still at the same time like like being as standard as possible when the standards make sense oh there we go oh okay now leave the grid disappear how do I view future social networks a lot is theia what's the future of social networks you don't know okay well I mean it makes sense that you don't know right that's why know you're building Firefly as of multiply so get like we podcast lights and Twitter at the same time right yep um that's yeah I think what do I think the future of uh of of this uh sort of uh sort of stuff is I mean I feel like like I feel like the next step in you know like things like forecaster and ones becoming more the success more successful is like they need to go further in offering features and functionality that like like offering some kind of benefits to people that Twitter does not offer and like I think it has a little bit of that because that you know like it has when forecast has channels they just have a different culture ter kind of people post but uh and then like you you you have this like the idealistic benefit then you have like different clients that are talk that are still sharing the same network like I think what I would want to see is I would want to see people really fol like asking the question of like well how do we extend these things so that you can do stuff that like you can't do on Twitter today right so like one example of this is like I think it would be cool if something like The ethereum Magicians Forum was just built on top of foraster and so you could we have you know like some kind of verge discussion where like people still interact with it through that interface but then at the same time you like will Bean would automatically become something that you could talk about through a forecaster Channel um so there's like things like that I think uh there's like this whole question of like is there something useful that the space can do in terms of the like Financial incentives side of things right and like I think I know like you know like social f as a category is like incredibly cringe and you know that's like you know there's been a lot of things that just that when we really make sense in the short ter right but that know I think there's still the open question of like is there a version of that kind of stuff we even compensating like content creators and that like actually get like make sense that it's something it can like really be stable for five or 10 years but I think uh like I feel like this space hasn't quite figured that out yet but you know if it does that would be really cool uh so there's that um the timeline of stainess oh that's I mean this is an interesting topic right because if this is one where like a of my thinking has been we shifting a lot recently right because I think the default path for a long time has basically been okay we have Petra and then right after Petra the next hard for is resered for for Ries I think there is in my opinion a good argument for like actually revisiting this and the argument is basically a combination of two things right one thing is that if you actually like ask you know what are the pragmatic benefits think that for which are supposed to provide right then one of them is okay you should be able to State without and like run a no without a Marg right but then there state is only half the problem the half of the problem with history and so we need 4444 influen it first right then the next question is well what's the underlying goal behind staking without a hard time the underlying goal is making staking more accessible well is staking accessible if it requires 32 e right and it's not 32 eth cost you know way more than AIC right so you know I have like this laptop here it's got a you know a ref archive node on it like the full two terabytes and you like actually you know like CH here helps quite a lot in basically you know like going into you know like the fancy electronics parts stores here yeah typ and like upgrading the 32 gigs of RAM the 64 gigs so can also run the 7D models on it and then also like adding a 4 terab drive on top of the one terab and like that stuff Al together cost like $400 so this laptop cost $2,000 right and then guess how much 32 e costs right well it's I it's back under $100,000 but still it's it's like 80,000 right now right and so like basically if you like talk think about underlying goals right then like basically get you know reducing the 32 e minimum is one example of like a different thing that you could focus on and so like I think there's a big question of like well maybe you know there's value on like delaying veral by one more for and like actually you like saying let's like prioritize other decentralization related tasks like a bit ahead right and like one example of that is like reducing the 32 we minimum another example of that would be on the execution La year side you know doing a account abstraction kind of right and like getting to the point where we actually have like some kind of ined abstraction for the validation step like this you know we can think about this right then so that's one part and then the other question is like I think there's a lot of like hidden costs that cors don't Rec don't like naturally recognize in con especially po researchers think recognize for researchers those right in like constantly rej juggling pieces of the protocol right like saying we're upating the protocol and then later we're upating to something else right because it's like every time you change a piece of infrastructure the windy counter resets to zero right and like there's lots of applications that has to upgrade and then the possibility of core infrastructure changing means that more applications have to have upgrade switches and then more like more solar sakers have to upgrate infrastructure there's like all of these hidden costs that like harm if your's ability to be this like you know incredibly neutral Global sum layer and and so if we want to optimize for that then like the challenge with B is that eventually qu of computers are going to come and quat of computers are going to like make anything based on what the curves notable anymore right and so the other path that we could take is we could say well how about we SK work we going to go straight to starting by has functions right like basically again you know and within that design space there's like a trade offering basically you take either you take an aggressive hasch function and then you do a start on that right like it take beside it like you can do 500,000 hashes a second of laptop already right and or you take a more conventional hash function like BL three you can like you can prove like maybe 30,000 hashes in a second right but even still um it's like you optimize that a little bit more or maybe you know if you kind of do some kind of meaning in the middle of those two approaches and basically yet within the three or four years we can get to the points where like we were actually feel we would actually feel comfortable also flying the ethereum state TR right and like if we say we aify ethereum St Tre we actually make a hard commitment to the community that like we are never changing this again unless there's like actually sound like big security wise one then that would make applications in all comfor with like for example doing applications that like d within l2s that directly be groups that are on one right and then like I you know there are other ways to do this what you could theoretically have a briefing pile that like upgrades if the St Tre upgrades and that's something that I was pushing hard two years ago but then it turns out that actually doesn't work the reason why is because there's lots of applications that do these State proofs inside of a store right and so like basically it turns out that like well actually the biggest gain do basically just come from like actually also State TR right and so if instead we prioritize other things and then we instead aify the state tree and like we do all the stess this stuff on a somewhat more delayed schedule then like there are benefits for that right I think it's like really worth it for the community with like can the different L of these but yeah I think uh I feel like right now it's kind of on standby right like there's there's people contining to research know BDF and still an open question of like is there a BDF that we can actually become confident enough in and I think if we do discover that then like eventually well I think there's a good chance that we put it in um but I think right now it's like one of those might open question marks yeah I I saw delayed encrypted and I wonder if it is used to resist or yes so I guess I VI to the cryed m as being like in the same category as like some of these like multi proposal solution right in sense that they try to like remove all like many kinds of any strategy from the space of right because like they basically say like everyone is going to publish transactions that are encrypted until like basically after they were alled to and then they get encrypted right and so like this ends up soling a lot of you know private order flow incentives and all these different things right um I think I the one big challenge well that stop is technological right which basically that like delay encryption is like kind of almost the same technology as BF like it's not quite the same but it's in the same space and so we need to have a construction that we're actually coment in uh I have my question is more like Rel to DP VM or so my first question is uh you you wrote a post about to the last B yeah yeah so I would like to hear more insight about this this this idea yeah and yeah sorry and the second one is about the proving system you mentioned U uh stock in many of you post in the Tre and we all know it's SP resistance but there's a there's a major argument in the field is uh stock is not is security is not fly verifi so I would let you know you you'll s out this security okay um okay so on the multi prooving I think so the idea is right that ethereum has this multili velocity where we have multiple implementations of all of the pieces of the ethereum stack and so know that both like produces power concentration in one group of Developers and it also makes sure that uh if one of the clients has a bug then like that bug doesn't take down the whole network right and so I the question is like if we start incorporating ckvm into ler one and like we start to make big parts that dependent on later one then it makes sense to have multiple implementations of the CPM for basically the exact same reason right so that's like basically the that you know the philosophy there on the Stark security side so the way that this works right is that basically so the security of a stark depends on the number of branches that you do at every level of Ry right like the number of Branch the number of like different queries that you need and you get a veral branch for each one and basically what they what they have is they have an UNC like they have an unconditional proof that a stark is secure if you use a certain number of branches but then Starks that are in the wild right now they generally use two times less branches and like they have only like conjecture based aring notes or why this is right and so basically there's like this 2x gap between Starks as most people deploy them today and stars where actually can prove that they're unconditional the Gap is only 2x right and so if we deploy the m one I think we should just eat that fast and we should just B St St big and quick follow a question you know what is your expectation um like how many years the DK or can repace the current destion May so I think the uh the big milestone that I'm shooting that I think we are shooting for is that we want a ckm that is one of fully formally verified right so we have like a VAR strong formal verification effort that's like really starting right now that's been running for a few months and already starting to make quite a bit of progress and then two is that that gives proofs in real time right like basically right now zms generally take like 10 or 20 minutes to prove a block and like we want a zpm that can prove a block in like less than one or two seconds right now that does not mean proven one or two seconds on this right what what that means is it's okay to require like 300 gbus right well what it means is take advantage of parallelization right like basically split up any evm block into let's say 1,24 pieces have a different proof for each of those pieces and then have like a treaty of aggregation like proofs on top of proofs right and so then the time that it takes to proof to make that proof is only BL in right the resource requirement is s in the time is BL in um and what we have now is we have things that do one layer of recursion that can prove average E theum blocks in 15 seconds right so that's average case and not worst case and it's like 15 seconds is still like being 10 times more than we want but it's getting pretty close right like even 15 seconds that that's already good enough for client right so I think if we can get those two things done then we can get to the points where like ethereum clients not just validators but also even like like clients right like people like what I want is like I want a future where even like mobile clients and like metamask and like browsers and I am token they're not just trusting RC providers they're like actually fully verifying it right and like to get to that I think we need one is we need statelessness uh two is we need all of this Layer Two stuff and Med is know we need that fully functioning layer ones and GS for how long it takes to get there I think maybe I actually think we'll have tests for like we'll get there within that's one gu I think hold on I think Z EVMS are like the one technology that just keeps outperforming our expectations how how are you formally verifying corre construction uh so there's a couple of approaches so the way that all of the ones that I see work is uh so what they do is they basically uh they prove a much more minimal VM like generally it's like the two that I know about are KY and risk five right and then they basically formally prove an implementation of the evm in that DM and then they prove that it's equivalence to something like you know the pre-existing evm this right and then for how you do the smaller thing there's like there what you're trying to do is you're trying to prove that the CGA erization like actually matches the architecture the St right and there there are some approaches that are correct by instruction there's some approaches that like actually try to make a proof in lean so they start from an existing Al implementation they try to Pro it I'm like I I've seen some about so is question suggest by so um so in in order to achieve account extractions there's wallet that are willing to like signers from the outside world right and there's some Yankees that suggest we should in shine new crypto curves namely but it is like it must like intentionally uh avoid by he suspend that is like back door by some agencies so I'm wondering how you view this kind of potential conflict or like let's say what if in the future people want to further inine some lure crypto Primitives I think this is the reason why like I don't especially at L1 I don't want to like keep in shining application specific phtography right the thing that ideally actually wants to do is it wants to like inine some like minimal piece that's powerful enough that you can Implement other cryptography very fast on top of it right and so for that like the two existing efforts that exist right there's Max which is the modular arithmetic extensions which is like on top of e and then there's the SD e blue from Greg golden from many years ago one idea that I've been talking with the about Max is basically I think you should combine Max and together and basically make like one that just does modular operations on multiple inputs at the same time and do this like a way that's as optimized as possible and then if you do that then like that makes has functions like beside in much faster that makes to curve operations much faster like it makes lots of things much faster right like there's no good theoretical reason why it needs to be more of it like maybe two times sler than a pration right like I think even with our experience with free compiles so far there's like lots of free compiles that just don't get used anymore right and so like I think because of that there's like very little appetite at all want to keep adding more PR files but if we actually can add something that's like generic but that's still powerful enough then you can Implement both R1 and Quantum resistant cryptography and whatever else you want on lot then like that's something that couldn't yeah um what's your take um um economy security is a mean on the economic security of it's a of what sorry it's a b e ad oh the economic security of hasn't intern security is a meme oh is a meme oh oh that topic yeah um I mean I yeah me I saw some of those uh discussions I mean I think uh like basically uh the reason why I think uh like a lot of that is not true right is that I think a lot of the or sorry I mean like it's not true that Economic Security is just a mean is because uh a lot of the time when some of those discussions you it and I know this like people in you know like Avalanche and salana and other systems like some of them say this right they say like oh if something bad happens then we just slash socially and like the social leer figures everything out right is that like I don't think that we should trust the social that much right like I think there's things that we should that like there's places where I think like the Bitcoin maximalist approach just like doesn't uh like it doesn't because it doesn't it T them not trust the social L but like it actually does right and like I think that we can go to go scam but then I think the approach of just saying like hey let's like just give up and let's like assume if a 51% attack happens then like a committee of people are going to decide who's who's billions for the wor blank to bring right like that's uh like the idea that that could go well I think it's like itself very unrealistic especially if you think 10 years ahead like you think of this as again like major thinkers not being and like major for developers not being the like the openness and sship resistance in point etherum people but being know these very mainstream corporate acors right like if you imagine you know like Google and apple and Microsoft having like 40% stake and then if you imagine know they end up being on the wrong side but like actually they deserve to be slashed like you know they're going to come up with like very loud Arguments for why oh actually you know they need to be a big bail too big to fail or whatever right and so I think actually like trying to automate the processes of what happens in a 51% attack as much as possible is super valuable right and even having penalties is super valuable because even if it's never certain what's going to happen if if an attack uh like actually takes place the fact that there is a risk right the fact that there is a risk where like if you're and you know even my else and like you know that if you participating in a 517 de there's like actual running code that's going to burn your money by default if you do an attack like even that risk is enough to be a really strong disent devisor for for people right and so I do think that there is like that value at the same time I think the place where like that stuff does have a point is that I think the difference between like one million e of Economic Security in like 40 million e is definitely less than a l of people think right and so this is actually part of why like this this is part of the compromise that orbit right the way that it gets more decentralization is it basically says well instead of making all of the E slash we only going to make a relatively small portion of e like actually we get subjects slashing at any one% time right you have unique that different type of compromise one more question um is it possible to make uh the yeld Sol that yeah I mean I think the challenge is like if you give higher interest rates to solo state or higher rewards to solo stakers then like there's uh always that like the big guys are going to try to pretending 10,000 little guys right and like they're going to be able to get like basically the same rate um so the question is like how do you try to that the best that I've come up with is this EIP that we've like did some network analysis for and before with tonor which is this a approach that's based on correlation analysis right the idea basically is that like if you're a validator you penalize people's mistakes or if they occur at the same time as when other people are making more right and like this is something that we kind of do already within activity leeks but the idea is to like really do it for like basically every type of failure and there's some interesting like you go through that if you go through the E research discussions like basically what it ends up being is that like the status quo is basically that the most well resource actors they get slightly highs because they have like they much better technical setups but what you get is basically this correlation analysis could or like correlation penalties could reduce the like the benefit of being a large Staker by like something like I think three or 4X so it's like a a pretty big gain but that's like just from a a protocol perspective right I think if we start talking about like how you compensate for you know like this uh these other forms of uh you know various you know like T yield I think I mean to me actually in the most the most realistic answers there is just like some form of stage capping so that basically you know change incentives to Target the amount of St being lower than like half some other number of all the e basically try to encourage the the two different pools of people St and people like getting to like actually do different right like that's uh like that's kind of EAS so right the the challenge with all those other Solutions is that most of those projects they don't just want an assurance that you get slash they want like the ability to like actually take your money right like if you think about something like CD they want the ability to like actually take your money if you get liquidated right and so there's like this inherence trust aspect and that's something that like Source acers can't get but that is something that sing gools could get and so one other direction to try to achieve that is also try to have more types of decentralized staking pools that try to distribute Authority even more than what something like w was doing so that like that's probably the most thing you can do in protocol and then extra protocol I mean obviously again you know you can have like l2s do air drops Target Sol FES and then like they're playing a constant can of no game I don't know we have felt like big sers pretending to be solar sakers and like maybe kind of embrace the cat Andy that's just like part of the fun vum but yeah like basically try to give like solar stakers more of a kind of extra protocol that I think it's interesting right I mean there's like other fun ideas that you can do with the other one is like I think if you have a solo sing box then like you might as well read your set instruction like run your own local copy of 405 on it and so you have your like selfing so like there's like there's like lots ofsn like extra probable things that I think we still aren't fully post maxing out on either so I think it's some combination of both of those oh nice you remember the numbers yeah oh then we want want to finish a few more of the remote yeah let's see the remote let's see what the remote Declaration of we two chain I mean I guess like the question like integration like what you want out integration right because like I think like to me the the North Star is basically that like Willard two should be like experimenting today and optimizing for for separate goals tomorrow in those Dimensions where it makes sense for them to do that and ler should be that standardized where there Val this like ethereum wide shared share us their experience and so trying to know like continue to on the standardization side right basically that you know like starting with as the transfers Mo like client stuff then eventually getting into key store wollups and that's I think are like the first things to start from I think there's like also more of a social elements to this right which is like we've had lots of good interop events between ethereum CL teams but we haven't really had good interop events for wallets and we haven't had good inter for leader right and so I think trying to like experiment more with different formats is another thing that I'm only trying to do more of and I think this also an example of that yeah okay SLE w considerations sampling BL I probility samples consideration the etherum Foundation is a a research organization uh it's specifically s Sol as a protocol research institution it as a result it doesn't even have a w engineering team or does it have desire for example despite yes um some people believe the ethereum lacks the capacity to push um and there's um and there's even a consensus that if the effort to actively participate in the engineering efforts uh it would uh compromise the uh the mission of decentralization and no this is definitely like a really uh important discussion that's been happening within the EF which is basically like what is our relationship to in like these big infrastructure projects that that are like big and that need to be done and that have a that do need you know some high level of coordination and standardization but that's uh at the same time tried like 30 cents like some profit motivated incentive for it right and like I think two or three years ago the ef's approach has definitely been more on the side of like not not our responsibility with the one exception of client teams or for client teams we did like pretty heavily speed all of the client teams with some grants good now actually it's like a like most of them are actually quite self-sufficient and me least fund side defic which is pretty fascinating right but before the there this issue but then now you know we're obviously starting to think about these you know cross ler to standardization things and so like actually over the past couple of weeks I've been having discussions with people in the EF of like how the EF could have like either in team or an independent organization and like Med doesn't even matter too much which one that's like much more specialized than like being closer to the user and and like interacting more directly with people that do that are doing l2s and wallet so like that's a type of discussion that we had another discussion that we' had is in the context of Z idty stuff right like there's a I think a desire to like really take some of the work that the PSC team has been doing and like basically go from saying you know PSC privacy scaling explorations to you know PSP privacy scaling Productions not wasent world and like basically saying like you know we've been talking about like digital democracy let's like actually have like CP that people somewhere actually use and uh the question there is like is PSC going to rol itself into being a know much more production oriented organization or are we going to take a like this kind of like arms length approach and so far I mean like we're still navigating that and actually the answer so far has been some of both right so like with CK email for example like that's something that PS has been doing and that's something that DC has been like pushing closer and closer to like a production standard quality and that's something like a bunch of flets are already starting to incorporate as a like as a social recovery op right so that's like one of those you know like examples of like actually getting like us getting a Clos to like actually building stuff that that people do rather than just doing researching demos but then there was there's other places where there's like specific projects that need to be done but then it just turns out that like there actually are high quality de schemes that like already came together for some other uh purpose right and if we can just like work with work with them and Empower them and like especially if those schams or values aligned then like the right thing to do is not to like try to eat them you know some kind of app high but to like actually get you like find that ways to like have that like collaborative relationship with of cross communication so like this is definitely something withf like very much like in the process of C yes oh how okay see okay what do I Vision the future of solar thinking look like I feel like I've talked about that um do I see reing as complimentary to ethereum or something so should approach with caution um I mean I think like the caution side is basically yeah what happens if there's a like the type of reaking that actually involves putting lots of assets at risk right as opposed to like I where there's like one component this the other component is tokens actually risk like that type of thing I'm like actually much l worried about right like like if we actually start to see these like very large scale I like re hypo type uh applications then you know there definitely is a potential know risk unknown unknowns if they D get too large but like right now I haven't seen one that's like that that feels like it's you know like has natur room to expand to be large enough to be a concern right because like right now we have 10 there 30 million e that's state right and like realistically I feel like all of the again lers to Sol here combine like I probably Define with like a few million of reaking oraking security but you know obvious that could that could change and find right so I think focus on like take application based approach is that I feel like the subs are going to induct the appliation specific um but you I think know the value that can provide definitely compliment um so you the impact of the eum after the adoption of theum TF okay so in uh in Paris there's this like joke that I heard once which is that what's the most beautiful place in Paris to get a the answer is why is it most beautiful place to because it's the only place in Paris we get a few have to look at the and so the like the like here is that I think the biggest benefit of the ETF being successful for ethera my think is like we don't actually have to like think about the freaking ETF anymore right like like like honestly like I feel like it's like it's had this like it has this kind of like resource curse effects on a on a community where like instead of the community optimizing itself for either theistic values of you know I open L this or Miss L this said resistance or de sensualization or the pragmatic value of like a like let's that make some of these things actually reach mainstream users in any application category payments like it basically turns the if eem into like very like small be kind of characters just like look up to the GS above and like please us right like Del of that's done like I'm actually so happy we can actually you like focus on the improving things like you like distal Sy at the same time like getting much more serious about like like mainstream application and S Bo questions [Music] thank you