📺 Super Scaling with StarkWare | Uri Kolodny & Eli Ben-Sasson

hey bankless nation it is time for another estate of the nation super excited about today's episode this is super scaling with starkware we have two the two starkware co-founders on we have ellie and yuri of starkware we're going to get to that conversation in just a minute but david this is the time to bring on starkware i mean this is uh zk rollups rollups in general are uh are starting to really take off i think in a big way starkware is rolling out some crucial tech why is now the time to have

starkware on to tell us about the scalability strategy for ethereum and blockchains in general yeah i recently tweeted out that layer 2 is here it's just not evenly distributed yet and some components about ethereum are super scalable dui dx is doing more volume than coinbase immutable it just released it's a gaming powered gaming focus nft platform and it's already like throwing out the the transaction throughputs the scalability that we all want to see on ethereum is being built in these different corners

yet that is the first version of starkware's products dark x stark net is also coming soon all these conversations are are happening all at once and definitely in the moment of where like crypto twitter just absolutely blew up over ethereum scalability it's a fantastic time to bring on starkware to talk about how scalable zk roll ups can actually be when put on top of a decentralized permissionless settlement layer that is ethereum yeah i feel like uh it then the knowledge isn't evenly distributed either right it's like we're doing our part but a lot of this technology is uh is hard to understand

and so part of what bank was trying to do is break this down into mental models we put together the uh the modular modular blockchain podcast you wrote an article about that david would refer listeners to that if you want to get a high level overview but we're now on a journey towards like unpacking these zk scalability type solutions and starkware is definitely one of the leading ones one of the ones we are most excited about in the space so gonna be a fascinating conversation stay tuned to for that also some quick announcements we dropped a podcast with visa visa is

going all in on crypto in the crypto space other fintechs are following that was an episode we just dropped on monday david any uh sneak peek on that app uh not only visa but also anchorage which is really the emblem of the d5 mullet you got visa in the front you got anchorage in the back what is it like to completely redefine what it means to have fintech fintech based on crypto rails not central banking rails so all of that and more in the podcast that came out yesterday on banquets there you go d5 mullet thesis right so the mental

models just keep flowing out of bankless guys this is uh this is a crucial stuff for you also we need to give a shout out to opolis they're doing some cool stuff these days they are sponsoring this message i tell you david i haven't i've heard i've heard this so often like somebody telling me i'd quit my corporate job my boring corporate job in a heartbeat but i can't because i'd lose my health care this is of course the us i know other countries don't have these problems but this is how they keep you a wage slave this is how they keep you a

member of uh the corporations right you might want to go where your heart is go towards a dow go where the opportunity is but you lose benefits opolis is an organization that services dows that helps solve this they do payroll and they do benefits for daos and for self-sovereign workers we think the future is moving from the gig economy to the ownership economy towards self-sovereign workers and opolis provides provides benefits for those anything else you want to say about opolis david yeah this is really the

infrastructure that we need to go fully sovereign and not only when you sign up for opolis can you start to have health care with your other dow workers right these are all the same health care albos is like a co-op you guys all get to like choose your health care together we all get to be insured together and also rates go down as the organization grows up gets larger but if you sign up for opus you get a thousand work tokens and a thousand bank tokens if you sign up by the end of this year uh and so the work tokens is the tokens of the opelus network again it's just just like every

other dow if you use it you become an owner in it uh and so opus uh making sure that all of the infrastructure needed to be workers in the dow decentralized world are there so that we can actually sustain this ecosystem going forward guys there's a link in the show notes if you want to get plugged into that bankless dot cc slash opolis david i want to start with the question i ask you in every state of the nation what is the state of the nation today sir i've done this one before but i'm doing it again today we are building and the starkware team has been super hard at work building out zk tech for

everyone on ethereum to use as we go through a weekend of drama and fud uh the only answer to really any of that shenanigans is to just keep on building uh and that is exactly what we are doing here on the state of the nation we are building today i want to find out from the starkware team if they are abandoning ethereum my guess is their answer is no but we will get into that in just a minute before we do we want to thank the sponsors that made this episode possible arbitram is an ethereum scaling solution that's going to completely change how we use d5 and now it's live with over a

hundred projects deployed gas fees on the ethereum l1 sucks too many people want to use ethereum and it doesn't have enough capacity for all of us and that's why teams like arbitrage have been hard at work developing layer two solutions that makes transactions on ethereum cheap and instant arbitrary increases ethereum's throughput by orders of magnitude at a fraction of the cost of what we are used to paying when interacting with arbitrary you can get the performance of a centralized exchange while tapping into ethereum's level of security and decentralization that's why people are calling this ethereum's broadband moment where we get

to add performance onto decentralization and security if you're a developer and you want to save on gas costs and overall make a better experience for your users go to developer.offchainlabs.com to get started building on arbitrum if you're a user keep an eye out for your favorite defy apps building on arbitram many defy applications that are on the ethereum layer one are migrating over to layer twos like arbitrary and some are even skipping over layer ones and deploying directly on layer twos there are so many apps coming online to arbitrarium so go to

bridge.aubertrum.io and start bridging over your ether or any of the tokens listed and start having a d5 experience that you've always wanted living a bankless life requires taking control over your own private keys not your keys not your crypto that's why so many in the bankless nation already have their ledger hardware wallet which makes proper private key management a breeze but the ledger ecosystem is much more than just a secure hardware wallet ledger is the combination of the ledger hardware wallet and the ledger live app and if you're used to seeing all of your crypto services and favorite defy apps

all in one spot ledger live is where you want to be not only does ledger let you buy your crypto assets straight from the app but it also hooks into all of the defy apps and services that you're used to using ledger live you can stake your ethan lido swap on dexes like paraswap or display your nfts with rainbow you can also use wallet connect inside of ledger live to connect to all the other defy apps that keep coming online defy never stops growing and the ledger live app grows alongside with it so click the link in the show notes to see all of the

defy apps that ledger live has and stay tuned as more apps come online and if you don't have a ledger hardware wallet what are you even waiting for go to ledger.com grab a ledger download ledger live and get all of your defy apps all in one space guys we are back with the starkware team pleased to introduce you to ellie from starkware co-founder of starkware also yuri co-founder of starkware gentlemen great to have you on bayclass how you doing today hi guys thanks for having us

well we are super excited to uh to dig into everything that starkware is doing in this space and we're excited because um this is really this is really kind of the scalability answer i think to crypto and to blockchains and and you guys are are really delivering it and ramping up the delivery of it so we're going to talk about starkx dark net everything that you're delivering but first we want to congratulate you on this massive series uh c fundraise so there's a 50 million dollar fundraise a 2 billion valuation congrats on that how does it

feel to be on the other side of this fundraise well it's uh you know it's uh the the series c was done with a sequoia they've been our investors and close collaborators since our series a going back to 2018 so we've known mike and the sequoia team well and that made for a very simple and uh easy process so regardless we're happy to to have this sort of uh milestone behind

us i want to say that it's better to be on this side than on the you know site before that of course of course you guys there are so many different things that we want to dive into with uh everything with starkwear but first uh you guys have a lot of things that are named stark something and so at the start of this podcast we want to go in and actually label and define these things because we're going to be referring to them more and more throughout the rest of the podcast so there's starkwear there's starkx there's starknet can you guys define and delineate between these three

things for us sure so first of all starks start starts were co-invented by uh two of my co-founders uh michael bryaptev who was our uh chief architect uh and got his phd under ellie uh before uh co-founding starkware with us and with alessandro chiasmus they're co-inventors of starks that's the zero knowledge protocol that we're using in our software stick uh starkx was our first product and this

is a permissioned uh standalone customizable scaling engine that runs over ethereum and this powers immutable x and surveyor and dydx and diversify and suitable power seller as well starknet is our permissionless decentralized ck roll-up and the alpha has been live on a public test net since mid-june and is going live on mainnet ethereum before the end of this month

okay so just to reiterate that starkx are these multiple scaling chains that are app-specific right so dydx is using an app-specific zk technology uh platform which we call starkx and also diversify is also using this and sober is also using this uh and immutable uh but then starknet is a maybe a little bit more what users might be familiar with when it comes to like optimism or arbitrarium one where there's one canonical chain that is permissionless

to build on and everyone can build on that that is not let yet live but has been in alpha since since early june and maybe we have some announcements coming soon about its uh release date maybe is all of that correct yeah yeah yeah yeah so as as as i mentioned it's been on a public test and since mid-june it's coming to humanity ethereum before the end of this month i should just uh just one my minor correction uh the stark x instances are not chains they are essentially uh

applications services scaling services uh powered by the starkex cloud service running over ethereum okay let's unpack that a little bit they're not a blockchain so where does it how does it why is it not a blockchain where the differences lie and how does it actually achieve scale elliot do you wanna um so um i think what defines a a blockchain i mean there are several things it's not there's no formal

definition but it's sort of uh this um uh it's very permissionless it's very you know everyone can come submit transactions and uh you know it's something that basically the public sort of uh runs and uses and um star kicks um is is as you said the line of products that each one of them is sort of a back end that services one particular customer that is another business that is customer facing

so dydx it's an exchange that faces customers and it uses a technology stack that is starkx it basically uses the zk start technology in order to compress computation and achieve computational integrity without any trust uh but for trust and math and trust in in the blockchain so it's sort of an interface and a settlement layer that's very specific to a customer um that's starch so i i think it would be a stretch to call it a blockchain uh it is more an

interface in a settlement layer and sort of something that boosts up the um capacity of some other blockchain and in all of these cases it's ethereum a compression service if you will yeah all right that's okay a compression service right so you guys talked about uh said the word cloud who is actually operating the compression service is that you guys yes for start x it is correct okay net it will eventually be a fully decentralized network where anyone could

function as the sequencer improver on the network okay hopefully including you and ryan stark is starcx is this technology that you guys have in your geyser's own cloud that that you guys run and you guys allow people like dydx or so rare or immutable to tap into that service yes okay and then start net is that same sort of service but more streamlined and just made more available

for everyone and rather than have it being on one single cloud it's is it is stark net an actual blockchain uh it's closer to it and there's another distinction so starkx is sort of crafted towards a very small number of uh special purpose use cases that are extremely important for instance uh dealing with uh a massive amount of uh payments you know very high tps like for a payment processor another example is a massive amount of trades or perpetual uh

swaps and positions or minting of nfts so it caters to uh things where you have very high demand but for very specific functionalities now start connect in addition to being permissionless um it is also a platform that is turing complete and universal very much like ethereum so it means that instead of us saying okay you can either you know you can do one of four or five things that our customers really want like you know minting trading transferring blah blah blah basically it

says okay anyone can write any kind of logic for any purpose they want it could be generative art it could be you know compressing many votes it could be some game design it could be a whole number of things that we don't actually know what they are we basically um you know the invented and then designed and offered tooling for writing um any smart contract deploying it submitting transactions to it so instead of limited functionality for these things that are in high demand it is this

general purpose uh framework so it's uh closer to a blockchain it's not quite a blockchain because it's a layer two right so i think blockchains are more layer one so just like optimism and arbitrarium i don't think they're described as blockchains they are layer twos we are also a proud layer too i wanted to add to that a very important feature that starkness has had for a while uh and will have on mainnet uh by the end of this month is composability and this is something that a lot of dapp

developers care deeply for this allows for you know for this money lego or nft lego or or a smart contract lego to thrive and have this amazingly accelerated and open development framework that that developers enjoy so much this this is super cool and i'm surprised as we're going through these names that tony stark you know wasn't involved somehow because this is super advanced tech uh but i'm sure you get that joke all the time um can we talk a little bit about the uh stark x first because uh i wanna wanna

just cover that and scalability now really with starkwear is kind of the conversation then i would definitely wanna spend the bulk of the conversation talking about start net because that is the next thing that is coming as you said you're you're releasing uh to maine at the end of this month so i'm we're close to the end of this month we are we are breathless with anticipation here but let's talk about starkx first because there is this impression in the space we just saw it last weekend okay literally last weekend i don't know if you guys you know hang out on twitter at all but the twitter world world was a

buzz with this idea that uh ethereum is not scaling today that there is no scalability on ethereum today hence the need for all of these other alternative layer one solutions and you know there was a lot of things i wanted to say but one of the things i wanted to say was of course it's scaling today right look at what stark x is doing look at what dydx is doing look at what diversify is doing look at what so rare is doing immutable x scaling using this compression layer

this this layer too on ethereum today now app specific for now but later uh this can be more generalized in in something like a stark net but can you talk about that a little bit because i don't think people understand or appreciate the the numbers here and what's actually being done across say these four the big four stark x applications today dydx diversify so rare immutable do you guys have any numbers here or any sort of metrics on um how uh what's happening how this is

scaling so far yuri why don't you start sure um you know dydx would be i think one sort of remarkable case study that we could point to um the transactions that dydex used to do on layer one uh each transaction composing of two trades uh a couple of markets only uh those were around 250 000 gas maybe 280 000 gas per transaction the transactions they're doing today on starkx touch many more assets they

they have the ability now you can put up collateral against many more positions and so it's a far more complicated transaction we estimate that equivalent transaction on layer one to it will have consumed about six hundred thousand gas per transaction those transactions today in production on star kicks consume sub 500 guests not 500 000 guests sub 500 gas per transaction wow now we're talking about this this isn't a demo and

this isn't uh like a nice sort of case study that we ran for an hour that brought down this is day in day out the past 24 hours were at 10 billion i believe trading volume for a dydx at the end of another one of their fabulous epochs uh cumulative to date over 230 billion bias and boy uh dollar settled on ethereum uh star x in total over 55 million transactions dydx is probably around

half that in total so a very very dramatic scaling today already on ethereum i we our estimate at 600 000 gas that's about 300 000 gas per trade that's roughly uh 3 tps would have consumed all the ethereum's layer 1 capacity okay we're doing on a very regular basis 15 times that on this on on star kicks yeah yeah this

is just uh show what we're talking about so this is the tps the effective tps that we're seeing i hope you see my screen share this is the effective tps over the past 24 hours so you see it's like on average i'd say on the dydx system so this is like i'd say on average a tps of seven uh with peaks at uh you know 13 tps and again as we just mentioned the effective tps if all of the ethereum would have been consumed for just uh settling

these these trades uh ethereum wouldn't wouldn't be able to to settle this many it can handle a tps of i guess one of them yeah so i i mean at its peak basically dydx is uh another ethereum right oh no it's it's multiple it's a multiple experience wow wow that's just one so that's just one of our systems and it's consuming on a regular basis you know over the past 24 hours we haven't looked at the numbers but it's gonna be less than one percent of ethereum's gas while

doing something like five to ten sorry 5x or whatever 2 to 5x the capacity of all of ethereum and consuming one percent of the gas yes and i want to get to so rare and immutable in a second too it's like but but but what we just said i mean that is scalability right this is why i feel like there's such a narrative mismatch in in in the market in crypto twitter these days and it's uh it's a very bizarre world we're in right now just just to recap there are multiple ethereums perhaps two to five ethereums

worth of capacity that is being settled on ethereum by dydx and what you guys said is less than one percent of ethereum's capacity so two to five ethereums being compressed down to just one percent of ethereum but i i want to emphasize that we're not supporting two to five because that's what we can support we're supporting two to five because that's the demand right right we can support substantially substantially more multiples of that easily today and so when i said uh

that would probably be another i don't know how many but like another five to ten ethereums uh you know we meant it i forgot how many million nfts over the past uh half year probably again that would have consulted i think about about 25 million nfts are we talking about so rare or immutable now or both of them both of them both of them cross both of them okay give us those stats again so we don't miss about about i think maybe 24 25 million nfts minted to date wow you ever take a few million here

that's a few million so what's the start of this show and i said that at the beginning of the show that scalabilities here is just not evenly distributed yet this is what i'm talking about dydx hyper scalability it's just uh people are expecting like this thing to be happening like ev everywhere equally across the board but no that we are getting hyper scalability uh like you know sub one penny transactions uh in very specific uh examples and these are applications yeah and these yeah and i think that starcx is bringing

well i wanna stress this these are production systems so both surveyor and immutable are working in uh valydia mode where they chose to have an off chain data solution starkx employed as a star off chain data solution uh because they wanted reasonable gas cost per per minute uh we are minting batches of 600 000 nfts at sub 10 gas per minute once again 10 gas not 10 000 gas wow that's incredible yeah and i think

bankless listeners will be familiar with what a volidium is because we've talked about that we did an episode on modular blockchains that kind of talked about uh and touched on all of these differences but it's basically modular blockchain if you kind of break it up consensus data and then execution right you're talking about valydium uses um not ethereum for its data layer uses ethereum for its consensus layer but it uses something else for its data layer and a volition of course is when a chain head kind of has the choice an application that's the

choice um i want to ask about this because this so this is a question incredible extreme scalability super scalability as we said is happening right now are there any trade-offs here refresh us on this with we're still camping on starkx right now but what are the trade-offs here with starkx from us you know decentralization perspective ellie what would you say yeah there are actually many many trade-offs that are behind the scenes i'll just mention a few of them so if you want

latency to be smaller meaning so what what would what would entail so right now we're basically taking a very large number of transactions and you need a very long time for them to accumulate and then you're producing one single proof for that so you need also a lot of computation to get that thing so what you get is you're increasing latency which is bad but you're decreasing and minimizing amortized gas cost per transaction because the larger

the batch that's basically what the magic of starks give you the larger the batch the lower the amortized cost per transaction will be okay so one very clear trade-off is that as we grow the batch sizes all things considered latency grows but amortized gas cost goes down another thing is that um as you um yeah i think that's probably the most important trade-off

there are other internal trade-offs in our system for instance if we wanted to keep if you kept the same batch size and latency in everything but you actually wanted the uh on-chain proof to be smaller you would pay with off-chain computation because there are various mathematical parameters you can play with there and that's another trade-off that we have at our disposal you know buy a bigger machine have even more latency and even for a fixed batch size you can

decrease that so i there there's an obvious trade-off in the context of data availability if you want to operate in roll-up mode your data is on chain it is no doubt more secure it is no doubt more expensive uh the beauty of the volition solution which is coming online in in a matter of weeks um is that this will uh this decision will be handed off from the app to the user and so the user can decide on a

transaction by transaction basis if they care enough for security to pay that premium on chain data will it literally just be like a drop down menu where it says here's my transaction i want it stored on chain or i want it stored off chain it'll be just that simple and here are the costs right and it'll have two different costs associated with that yeah that's pretty that's pretty fantastic indeed ellie can we go over that latency trade off again so it's like i kind of understand that but like what is the trade-off that an individual

user is making with latency right it's like if i have if there's higher latency how does that affect the i guess security position for an individual user does that just mean things get settled to ethereum less frequently so there's a longer window of time where they're potentially vulnerable that's a terrific question so uh currently on starkx this is very much abstracted away and the risk of this window is taken by our customers you know dydx

diversify immutable and so on now what does it mean it means that uh you as a user and and i'm assuming that many of your listeners have used some of these systems the ideas so they probably noticed that they have instant finality and even if they want to basically go you know retrieve all their funds on chain they get it very very quickly now this does not mean that the relevant stark proofs happen with the same finality but what is happening is something like this dydx which gets

these orders and settles them and matches them uh and calculates the positions knows that once it settles something or did something it basically now goes into this uh stack or queue where which is controlled by dydx and uh you know it will take some time for a batch to reach enough size for it to be convenient to put a proof and then for the proof to be generated and um and accepted on chain and currently for dydx for instance with its very large proofs

this this is a number of hours but the user doesn't uh experience any of that what's happening is that dydx is essentially assuming whatever risks are associated with that and i think they made a wise choice the risks are relatively small because um they know that um you know that something that has been settled will appear in a proof and that approve will be accepted on chain and basically they know what the next state of the so it's a little bit like i think in traditional

trading a lot of you know settlement is t plus three but a lot of exchanges will will sort of assume the risk for three days with a case of uh uidx and other customers it's a small number of hours and again it's in their control for instance dydx could decide that it wants to close a batch every whatever two minutes and then the proof would probably be much quicker to generate and would go on chain much quicker so maybe the latency from dydx's side would not be

hours but minutes but then the cost would be that the amortized gas cost per transaction would be higher because the amortization is over a smaller number so most of our customers have chosen to basically crank the knob to maximal batch size minimal amortized gas cost but also maximal latency which is a wise choice switching but when you when a bunch of more users come and deploy more transactions at a higher rate

you guys can actually uh increase the rate at which they settle to ethereum without also increasing the per user gas cost is that correct for sure and we can do that thanks to uh a remarkable technology called sharp which is which stands for shared prover and this is a technology that's in production today serving at the moment so rare and immutable and diversified and hopefully dydx will be added soon to this system and sharp basically