Starkware, L3s and the ZK Rollup Future | Eli Ben-Sasson, Uri Kolodny

Transcript

this episode is brought to you by Circle the issuer of usdc which hopefully as you all know is the preferred stable coin of digital Natives and crypto natives with over 1.5 million holders globally you'll hear more about usdc later in the show [Music] welcome back to another episode of Empire you got Santiago and I are co-hosting this one we've got Yuri and Eli from Stark where two co-founders of starkware Welcome to Empire Yuri and Eli thank you Jana thanks Santiago thanks for having me hi hi happy to be here yeah happy that you're here awesome guys well I actually want to spend the first like five to ten minutes relatively high level on scaling and then actually get pretty into the weeds of this this might be one of the more technical episodes that we've done

in a while um I think the first place I want to start is like there are two ways to scale a blockchain right the first way is you can make the blockchain itself have higher transaction capacity uh they're obviously problems with that the second is that you can actually change the way you use the blockchain right instead of putting all the activity in the blockchain itself I just want to get your guys's take on like why that layer 2 model is better um so maybe we could start there basically cloning a code base under you know uh that has a certain you know architecture and scale will give you pretty much similar scale or you know it it maybe initially it won't be congested um and then you'll think that that you

have much greater scale but then uh the same limits of the technology will kick in so friends instance polygon is essentially the same architecture as a that of ethereum but with roughly 10x uh the rate of blocks so you expect you'll get roughly 10x the throughput and then you'll be blocked by the way even this 10x comes at a cost it's not the same architecture it's the same software stack and it's it's six-fold the block frequency otherwise it's identical but you'll get uh so you you know you put in something you know you crank it up six or 10x you'll get six or 10x and uh then it will get stuck and uh you know then you can say oh let's uh let's have uh

you know smalligon and that will be 7x and same thing and you'll see it again and again but this is not a very good way to get security uh by the way already here with polygon what you're I mean what users are paying with is that uh you know the level of trust and decentralization gets lower and of course you repeat this this uh uh I mean you'll get what you what you uh what you expect you'll get you'll just get the same phenomena again and again so what starknet offers as an L2 and a very interesting L2 one based on validity proofs is a very different technology and it is one that uses the very same good old L1 which is ethereum but much more efficiently it uses it as this very

scarce and important and trusted resource that can assert Integrity over a vastly greater amount of computation even without processing that so this is a surprisingly different and far more efficient way of utilizing blockchain it's a very different technology and it just happens to scale much better and with it you don't get 10x the throughput you could easily get a thousand twenty thousand fifty thousand and a million x the throughput and as we'll describe later on these are not just theoretical limits this is the way our technology is operating today you got we're going to get into things that Eli I know you are excited to talk about things like uh

computational Integrity delegated accountability all of that kind of fun stuff but we're gonna have to dump some of this down for my brain is not as big as your guys's brain to kick this off so I want to uh like I want to make sure I understand that correctly uh so it sounds like the main challenge with that technique like the the other L1 technique is that blockchains with these bigger blocks are more difficult to verify they're likely to become more centralized the only solutions to that are really to have like increased the efficiency of the client software or to use techniques that like like things like sharding uh which kind of allow the work to uh of building and verifying the chain to be split up among amongst many nodes that's

the only solution on the other L1 side and that's scalable but not that scalable is what you're saying yeah yeah the state of polygon today is is a clone of ethereum so it's not yet as congested because it's not yet as popular but if you have the same throughput or 6X the throughput it will experience the same congestion and high prices of ethereum and this is something that has already happened when there were uh spikes of congestion on on polygon for instance so it's just uh the same thing and I would add to that that out of the gate far less decentralized right far fewer nodes can run at that capacity so so the the trade-off is well

understood you can increase the uh capacity the throughput that individual nodes can process by doing so you're excluding smaller participants from the network and you're uh hurting decentralization that there's no no new physics involved in in that trade-off and there's a limit to that in the end there's a limit uh you know you could go to one computer trusted computer and that's known as you know Facebook the Visa Bank of America uh Citibank model and it works very well but it's just you know it's not blockchain yeah okay so we're doing some process of elimination here so we've eliminated the other L ones now we get into the L2 model right and so with the L2 model there are two I mean again there are two main models that I know right we've got

uh optimistic Roll-Ups and we've got the ZK Roll-Ups so a lot of folks have heard about things like um like optimism and arbitrum have been in the news recently optimism to their token thing arbitrim has some interesting stuff coming out or they launched this thing called Nitro right so you guys have chose so that they do they do what's called optimistic Roll-Ups you guys do ZK Roll-Ups is my understanding so can you guys just get into that model and like optimistic Roll-Ups or ZK Roll-Ups what's the difference wait do you want to take it sure um optimistic ropes basically rely on crypto economics to uh allow the network to scale so instead of saying let's compute all these transactions on the blockchain itself you said let's

move these transactions and compute them off chain in a given party can compute those transactions and um in the process of doing so Stakes a certain amount of tokens um in order to uh to sort of ensure that it computes with Integrity other participants observe its uh conduct and if they can identify a fraud and that's that that's this is why it's called fraud proofs they can uh submit at least an arbitrum arbitrum itself can submit such fraud proofs with optimism uh the fraud proof mechanism is actually not enabled just yet um but in theory one could an observer of the network could submit a proof of a

fraud and then the entity that's computed uh that computation that includes such a fraud gets slashed so that's the incentive for uh for those entities that help scale the network to conduct their work with computational integrity okay note that in the process anyone who wants to monitor say Elias is one of those entities if I want to monitor his work I need to have computational resources that are in fact on par with his right if he's Computing a thousand transactions per second if I don't want if I want to be able to check his work in a meaningful way and keep up with the pace he's going in I need to employ the same amount of computational resources

now note that he's getting compensated by the network for every uh sort of uh honest piece of work that he's doing I'm sitting there waiting for him to to you know to uh to slip right to stumble and if he doesn't my AWS bills there's you know Steve still keep uh sort of coming in and I'm I'm getting no income in the process so there are significant challenges in the process uh there's obviously the challenge there of allowing me enough time to check Ellie's work okay and for that reason optimistic ropes have uh uh a window uh allowing to the detection of frauds that's somewhere between a week and two weeks okay and so that's the

window for finality uh in the context of fungible tokens this could be actually worked around um and and you could sort of overcome this by by uh you know essentially by by funding uh having liquidity provisioning uh fund this Gap in the context of nfts or non-fungible tokens there is not much you can do to work around that and I think that for that reason we haven't seen nft sort of uh get significant adoption uh with optimistic crops the approach we take is a fundamentally different one with validity proofs here is often called in the ecosystems you're knowledge proofs um what you actually get is a system that can progress from

State a to State B um in in our hybrid uh world we call it from a state a to State a prime that's much more sophisticated than saying State b um so um those State transitions can only be performed if there is a proof that can be verified that attests to the validity of that transition so it basically is a situation where the world before processing a batch of transactions is in a valid State and can only transition to a new state following the processing of that transaction of that batch of transactions if that batch uh has a proof attesting to the

Integrity of the transactions in that fetch meaning every transaction has been uh signed properly and and and abides by some business logic that the network is willing to accept Ellie you were you were uh I want to I want to just uh give a very clear uh example to the listeners of the difference between the capability of an optimistic roll up and a validity roll-up which is what Stark net is let's take a very simple example suppose uh you know Yano you and URI are transferring um you know one ether back and forth a million times because you're playing some game so you're bombing the mempool with just

transactions that move back and forth a single leaf okay a million of them now this should be processed and recorded on chain somehow by by the L2 okay you're doing this on L2 so an optimistic roll-up would need to put one million signatures in an ethereum L1 block which doesn't quite fit in and costs a lot starknet would need to put one Stark proof with no signatures on the chain and it will easily fit so optimistic rollups first of all do not have that great scale you can actually compute the limit of their scale whereas

with starknet in many cases it's very hard to even compute the theoretical upper bound on the scale okay and this is not a mere theoretical thing that you know if you play this game if you take something like the dydx system which in every block has a processing of many many many um Oracle price sticks each one of them comes with a signature so an in an optimistic roll up you would need to upload all of those price ticks and signatures onto the L1 for security reasons and in a validity roll up like hours you don't need so you get much bigger scale and also the security premises you know

what are your basing Security on is much much better and much more future proof and much well analyzed and understood in starknet than it is in um optimistic columns okay can I just make sure I understand this again um which is that so that one of the differences here is that part of a transaction is only used for the verification and is not relevant to Computing the state update then that can be left off of the chain but in an optimistic roll-up because that data would still need to be included on chain uh in case it needs to be later checked in that fraud proof that Yuri was describing uh but in a ZK role that doesn't need to get included on change do I do I get it it's nice to you you describe it much better yeah that's a

much better smack smell yeah smack on you know and and just to sort of uh sort of expand a bit on the example Ellie gave in the context of of Oracle service providers okay for both l1s and for uh optimistic Roll-Ups the Oracle Services whatever they are chain stacker and others uh the their consumption of on-chain resources skills with a frequency of reporting right if I report the price every hour every 30 minutes you know in the latter case I consume twice the blockchain resources it scales linearly with the number of assets reported meaning if I I report on ethereum the price of Bitcoin and schmidcoin Etc each new asset I consume uh the same amount of additional

resources but it also scales with the security of the system namely with the number of entities reporting that price right meaning if it's only Yano who's reporting the price and we hack in tiano's computer we can manipulate this price to fitziano in Santiago when in somewhat better shape if it's 2 in different entities we're in even better shape right because it's very difficult to tamper with but all those 2000 entities need to report their uh uh provide their reports on chain with monolithic chains and with optimistic Roll-Ups in order for the general public to be able to say hold on a second for this past period the median price of Bitcoin was whatever it is with a validity Roll-Ups we can in fact

prove to you okay that we've computed off chain the median price uh reported over the past period and that all these reporters appear on a white list of reputable reporterism blah blah blah and that the median is X and all we need to put on the blockchain is that X not the 10 000 or 2000 or whatever so in this context we decouple the security of the system from the resources it consumes on train and that's a very big deal Okay so okay so I think I think I get it now um so I obviously your eyes's argument is that ZK robes are much better I think I get the I get the argument I'm pretty sure um however like when you look at what's happened in crypto so far you look at

the two main scaling Solutions like the when I'm just looking at TBL like I'm not making any like judgments on who's better I'm just looking at tvl you've got arbitrim's got like two and a half billion optimism maybe like one and a half billion then then you take like a pretty big leap down you've got dydx is next which do idx is obviously built using Stark x uh and then there's like a and then it continues to drop off pretty significantly so I'm assuming the answer to why that is is because it's just a much harder Tech Challenge building what you guys are building and it takes longer but I'm curious to hear why like why uh these optimistic Roll-Ups have taken so much of a lead in the in the L2 space I think it's a very temporary and skewed picture I'll explain why first of

all you know even in the Star Kick system so many of our most successful um customers are in the area nfts which means I mean it's so rare immutable miria now joined and they have minted collectively millions of nfts actually tens of millions of nfts but if you look to uh you you know if you try to quantify those using dollars then then it doesn't show up as much or anything but these are very very meaningful and significant metrics that are not matched by any other system I don't know how many uh you know nfts are locked up in uh you know the combination of ethereum L1 optimism in arbitrum but I'd be very surprised if accumulated it's more than

one million over these three the L1 and these two l2s whereas we have uh close to or maybe surpassed already 70 million nfts again used in games like you know so rare is a company that was valued at close to five billion dollars just announced an amazing deal uh with the NBA these are you know valuable things that are happening and they're all happening on Stark X but you know they don't have the usdc or tether or if tied behind them so you don't see it in the metrics so that's one way in which which is skewed the other way in which it's it's skewed is that um for um you know safety reasons and because we're in Alpha we are limiting uh

artificially the amount of ether that can be locked and processed through the starknet bridge which means that uh you know you won't see as much uh tvl right now uh just because it's it's limited artificially as we're uh you know bringing the system uh to be more and more performant so uh yeah but if you look at other metrics just to mention a few you know um number of transactions you know um number of nfts as we said uh total number of transaction TPS and all these metrics we are more than technically the other l2s combined I I want to add another dimension here and that is that say with optimism and

arbitrum one can deploy uh uh solidity contracts essentially as is and let's put the security discussion aside but essentially can be deployed as is in terms of uh the developer experience in the user experience now our hypothesis and time will tell if this is in fact the case and then with with us uh uh we're not there just yet warp is a transpiler from solidity to Cairo that the nethermine team has been developing they have some big news coming up very soon uh but you know it's it's not as uh as uh the ux is it is not yet on par with uh arbitrum and optimism but our hypothesis let's say two years out okay

99 of the code that will run on starknet will be code that was written in this two-year period okay and in fact we'll never have run on ethereum L1 okay this will be native layer 2 code all this to say that if you think that two years out most of the code that will be run on there too is code that was originally written for layer one but then was deployed to Layer Two that's one strategy and then there's a certain set of tooling that needs to be built for that if your hypothesis is that the layer twos will actually serve native layer 2 code a couple of years out that's a very different ball game and we're preparing actually for the latter scenario firm okay that's interesting so a lot a lot of the code you're saying a

lot of the projects that we use in our products that we use in a year two years three years we'll have code that is written today and the code that is getting written today or next year is going to be written in Native layer 2 code and and I'll say why we think this this isn't sort of a sort of us pontificating on the meaning of life right that we we've seen this uh over and over again with uh with the projects and customers we've uh been working with right so so let's see dydx uh had the the good sense when you know when they started building with us the system that they launched on on Stark X not to come in and say and then we've told this story you know over and over again so so uh I I'm it's not to come in and say this is the logic we

deployed on ethereum layer one please you know deploy this implemented in Cairo which is the the programming language that we've developed for starknet uh I'll please develop deploy this as is on on Stark exters darknet instead they came in and said hold on a second let's start with a you know a blank sheet of paper right from a clean slate and say if we were to build a Perpetual uh contract trading platform okay what would we want it to do not what would we wanted to do on ethereum layer one what would we want what would we want it to do period and we started from that conversation and what resulted was a system that's not not only far more scalable than the one that they had running on ethereum

it's a far better Financial tool it's a far better Financial tool for Traders for users and that in itself is just a that was a very telling experience for us right that those developers who come in and ask themselves what is the right thing to build here as opposed to what is the right thing that could run on layer one ethereum would end up with a superior product let me I think this I was going to ask you guys about Cairo so maybe this is a good time um so like for context you guys introduced Cairo two years ago uh you just I think about a week ago introduced uh announced Cairo 1.0 first major release for for Cairo I think in like the last two years it gives Cairo 1.0 gives developers uh I'm just reading

from your blog now a safer simpler more usable programming language um I have a question about it which is all right so you so you mentioned warp warp is this like solidity to Cairo transpiler um are there limits does using the trans filer have limits versus writing directly in Cairo or maybe the better question here is like uh you mentioned that a lot of the code today is going to be written in like native L2 code um what about the but like we have to keep in mind that a lot of the big projects today are written in like not L2 code like the uniswaps and aves and and and the main like D5 protocols so like what a uniswap clone that is ported from eth1 transpiled via a warp via warp perform like how would that perform versus a

Cairo native uniswap implementation does that question make sense yeah it makes sense um yeah it it would work and you could go down that route um two things uh should be pointed out first um you you can never use code AS is that was designed for L1 as L2 even if you wanted to so like you know you would need to change things because the the storage space is different and so on and so forth so like uh you know you're anyways gonna need to change the code and um I guess my conjecture would be that even though it it it will work um you'll get more efficient scaling and

performance and even safer and more secure code if the main functionality of the smart contracts is written in Cairo especially in Cairo 1.0 so though you could do something I'll give an analogy suppose there's some really good uh you know functionality written right now in Python so could it be transpiled into solidity and run there I'm sure it can I don't know if it's it's very interesting think to think about like why isn't there any transpiler from you know rust or solidity sorry from rust or python or C into solidity right it's certainly doable why is Noah doing it and if you think about it I think the same answer as to why you know whatever answer you

give is the same answer as to why uh you know if it's a native Cairo will be the better way to go but you could take a package written in Python and transpile it into solidity yes and it will work but likely it's uh complexity in terms of gas cost is not going to be something you'll like the exact same thing if you take uniswap and transpile it into um a Cairo yes it will work but likely it's a consumption of L2 resources and also L1 resources when you run it that way will not be as favorable as if you uh as the case of rewriting and I just want to point out that on the stock net ecosystem there's actually a quite a

large number of folks who are actually building and writing amms uh in Cairo so I think they're going to outperform and someone there's going to be you know the the new year so yeah but though maybe the argument then becomes like well it's not going to be an amm model it's actually like Central limit order books we'll come to dominate trading uh because of the speed and and the low fees of stock where yeah even today there are no amms you know on yeah and conventional markets it's not because they no one thought about something like an amm it is less Capital efficient it works on ethereum only because of the decentralization and the gas cost limits but if you have refillable computation yes you're right it's going to go away that's good Ellie's Elias walking into a

religious War but I I want to I want amm's versus but but I want to say uh I'm not going to rescue him but I I want to say this though that that one way to think about this is so this is the metaphor I have in mind is that the folks building on ethereum and the unit swap the very talented engineers at uniswap would be one demonstration of that ability uh they have uh the demons the skill is to to build sophisticated logic as sophisticated sophisticated as they can that abides by the 15 million or at this point 30 million gas per block that's the exercise meaning there's a ceiling it's about a foot high and the question is can you with in in that in that room with a ceiling that's a foot high can you do somersaults and you can you do all sorts of acrobatics

and if you can you're a wizard and you really are a wizard right but now we're defining a very different environment right we're saying come out into this Green Field and do whatever you want now you can go out there and keep demonstrating to everyone that you can do somersaults abiding by this no longer existent uh one foot high ceiling okay but it's that sort of quaint at this point right this in in three years time this is like saying I I can fix typewriters that's wonderful but there are no more typewriters yeah that makes sense that's interesting I hadn't thought of it like that to yourself all this to say that that as as we see it the layer ones will be dominated in terms of of willingness to pay okay by the verification of proofs

that's what what we see layer ones will become the security layer so a proof a star proof that that's sent to the network to be verified in terms of of the number of transactions that it represents it already represents hundreds of thousands it will represent millions or tens of millions of transactions so the The Entity submitting that proof the prover you know submitting that proof to ethereum it is willing to pay far more than any individual player submitting a single transaction so we think that the the the individual individual transactions will be crowded out by market forces from layer one so I think Santiago and I are always trying to figure out like the second order implications of some of this stuff so like what I'm maybe thinking about right now is uh right now

uniswap competes In This Very like computationally constrained environment but maybe in a world of like L2 like in an L2 dominated World someone like unisot doesn't compete in a like less computationally constrained environment I I'm I'm willing to bet that's what's going to happen yes all right I'll add my ice cream to that yeah yeah so who does compete someone like someone who comes in and builds for builds directly in an L2 World summit someone we haven't seen before limit order I I think you see things that I'm aware like the um conventional model of uh right if you go Central limit order book yeah yeah I mean these are the things that work in the world it's not like uh

people cannot think about the uh you know economic theory behind something like an Autumn uh an amm it's just less efficient from you know slipping Capital efficiency things like that uh so so I think yeah though there will be amm I would put my bet on Central limit order books and as we recall it's not that Hayden started his career as a managing director at Goldman right uh he came up with with this very exciting idea and built this Powerhouse well I think he uh sort of showing up popping out of left field right yeah I think he borrowed it from a post from vitalik initially no there was also the banker protocol and then a few others in curve

and yeah but uh all this to say that he didn't grow up in sort of the the wild garden of of uh of Wall Street yeah and that's the beautiful thing about blockchain if you're a pro like if you're one of the uh I'd call it like the blue chip D5 protocols like a uniswap or Ave like you kind of know the like the old school D5 the tier ones what would you be doing right now how do you prepare for this like L2 dominated world oh I ca wait wait we can we can tell you what they are doing right or at least some of them Ave is building Eunice Ave is building Ave version three on Stark net okay there's a joint effort uh between the Avid Dao and starkware and

the uh another mine team uh uh put on this like you know what would like a a little sort of a Commando Force that's building this out frantically uh maker same goes right there's there's a joint effort with the maker Dao building stuff for uh for Stark net now I can tell you that in both cases they're sort of following the dydx Playbook which is to say they are not simply coming in and saying let us transpile line for line the logic we had on ethereum layer one to start no they're starting from a clean slate and saying okay this is a very different computational environment what do we want to build here and you know we're gonna I think soon

see some very exciting results in this regard all right everyone time for a quick word from Circle and uscc as a crypto user you know the power of stable coins dollar digital currencies that transcend borders banking hours and Legacy Financial rails while circles usdc has quickly become one of the most trusted and widely used stable coins it's simple people use usdc because of its composability its stability and its Reserve transparency and uscc isn't just adopted by a few of us D5 dgns and Dows and nft marketplaces crypto companies alike they all leverage usdc to diversify their treasury asset management and ecosystem-wide

composability the adoption's clear usdc has grown to more than 50 billion dollars in circulation since launching in 2018. we all have and we all will continue to take shots on our favorite volatile crypto assets obviously but usdc is one of the easiest ways to store your funds in a stable asset that can be used to send value around the world almost instantly it lowers the cost of cross-border payments it integrates into the growing ecosystem of crypto apps as a seamless trusted dollar digital currency usdc is a zero to one opportunity for the financial system if you want to learn more about usdc I would recommend you check out the recently published transparency Hub on circle.com it's a great update to

circles content on usdc it outlines everything from links to their weekly Reserve reports monthly attestations blog posts that are written by their exact team that highlight how and why usdc was built so it is really recommend it just go to circle.com backslash transparency to access it now let's get back to the show you got you guys have brought up dydx twice I just I mean I just have to bring it up like we just had an episode with the cosmos guys uh that just came out and so I mean the other kind of the other side of the coin like what what happened with dydx because they built you know half a billion dollars worth of value on Stark X I think it was and now they're moving to Cosmos um what's like can you guys just give

your side of the story there it's very simple uh first of all we we think of them very highly and uh I think it's really great the teams are um exploring and innovating we think that the way to scale is not by going to a monolithic chain but we you know respect the decision of our customers to experiment with that and we of course with the wish them the utmost luck um since uh you know this whole system is so great there there will be others that are going to use this technology if not the idx and I mean we'll see experimentation on both sides

right on you know a secluded Cosmos chain and on multiple instances on on Stark net and Stark X and you know the markets are gonna decide where where they think it makes more sense yeah I want to add to that that a whole bunch of teams uh just wanted to add to that that a whole bunch of teams uh saw sort of dydx going from about 20 million dollars a day in trading volume to uh essentially a couple of orders of magnitude greater trading volume on the starkens in Stark X instance that that they've been uh using as a settlement layer um and sort of our our sort of building very very uh rapidly to to try and

compete in in that space so so I think in that regard I think we are going to see a lot of experimentation there was some flag that dydx got within the ethereum ecosystem you know is how dare you guys desert the ethereum universe and and move on to Cosmos we we I gotta say as much as we are sort of very sort of strong supporters of ethereum and its ecosystem and its Community we we we take a very different approach as Eddie said we think that experimentation matters a whole lot and and uh so the The quasi-religious View that uh this software stacked that's seven years old this is sort of the the the uh really the the optimal software stack that

Humanity can produce in the context of blockchains we think is exceedingly naive that's why we're building layer twos now we're actually starting to also build layer threes uh but but in this regard we think that dydx going out there and experimenting in a very bold way is a very healthy thing for the ecosystem and we'll see how it develops I I I'm I completely agree I think it's I think it's almost comical the religious fervor that gets baked around a software stack right you would never if a SAS company left left Salesforce and they and they moved their CRM to HubSpot yeah you'd never be like this is blasphemy yeah yeah yeah pitchforks and all that yeah yeah yeah yeah let me let me ask one more question about uh dydx which is I think one of

the main reasons that they moved was to like uh what I read was like it was to maximize decentralization and I think that was in reference actually to the sequencer I'm just curious like what are the plans to decentralize the sequencer like would this be done via an auction I'm curious how what the approach is there we're now thinking about it we're in the context of uh um Stark net and it will be you know you know everything we do we we do very seriously and we lead the pack so I think we'll ex you know you can expect that it will be a very well thought well designed and something that you know doesn't break and makes a lot of sense just like all of our assistant I mean you know some of the listeners may not be aware of this but we've been running

Stark proofs which are acknowledges very Cutting Edge uh you know math and cryptography even among validity proofs uh these comes from the farther corners of that Universe things that everyone discredited and thought that just could never fly and we've put them into production with you know Turing complete language on fully functional systems that have been through multiple uh shocks to the uh to the ecosystem you know price surges and drops and everything and all of our systems have been functioning seamlessly we we we have this you know and when we say we're going to do something we usually deliver either on time or before and you can be damn sure that this will be uh you know

our performance also henceforth so there will be very meaningful decentralization and um you know if systems um want to have Stark's X instances that need to be decentralized those can be done as well there will be decentralization to the um to the need of you know the markets and customers and everything yeah so it's not really a technical um impediment I I want to add that in this context the sequencer is of course DUI DX right that's the entity receiving transactions from users and sequencing determining you know what goes into a batch in the sequence of transactions in the batch and stockware functions as the prover and those two functions sequencer

and prover are going to exist in a fully decentralized fashion anyone who wishes to serve in those roles would be able to do that in Stark net and in the not too distant future what about um I will admit I'm waiting into territory I don't fully understand here but um as of now there's no Mev on starkware because of the sequencer I again waiting into territory I don't fully understand but my understanding is there's no Mev on starboard because the sequencer how does that change once that's more decentralized nav is a Mev is a challenge uh it's going to be an ongoing challenge in every decentralized system so certainly once uh the sequencers of starkware are decentralized yes it will and there's

you know volume there and value then yes Mev is going to be an issue that we'll need to contend with uh are very practical approaches that you know we have so many burning problems to deal with um that were you know this isn't our first priority but you know many of the uh many of the proposals to deal or partially deal with Mev go through things that we're uh you know either things we suggested or that we're very good at things like using you know vdfs delay functions commit reveal schemes zero knowledge proofs so a lot of those are things that we are you know very familiar with and have been leading uh the pack on so when we get to it uh

yeah but but I don't want to give any false Impressions it's not you know in the next two uh releases of the code we're not claiming we're going to be dealing with Mev yeah cool um what are the big priorities I mean you guys right you guys have raised what 300 million to date give or take maybe a little less maybe a little more 270 or something but like what are we're very stable financially uh um and and we have a team at this point of uh say about 100 Folks at starkware there are more or less 50 people building out the starknet ecosystem worldwide outside of stockware and many many hundreds of car developers that weren't around six

months ago even so I I gotta say this is a for me every time I sort of I mentioned this it uh it's it's a it's a very moving experience to watch this thing sort of evolve in front of our in front of our very eyes uh but but the the next big milestone is car 1.0 I think this was mentioned earlier Cairo 1.0 uh is essentially Cairo uh rebuilt for a decentralized world so Cairo is the programming language that we introduced to maynet in the summer of 2020. that was done in the context of Stark X of our SAS offering and that's sort of a closed system in the sense

that there's a customer say immutable or so rare and there's a service provider which is Star Ware and there are certain understandings in that in the context of that relationship in terms of say denial of service attacks and censorship and all that you expect good behavior in an open network you can no longer expect such good behavior and there could be denial of service and there could be censorship uh enacted so this these are sort of two sides of the same calling right the denial of services is malicious behavior by uh people sending transactions and censorship is malicious behavior by by the sequencer or whatever um and so to protect against these two things and and in addition to that to make Cairo an even increasingly it's

Cairo started out as a low-level language and over time became increasingly higher level so car 1.0 is is the language rebuilt for a decentralized network and made even more higher level than than it's been to date um so that's the next big objective nice what about um let me ask about a couple other objectives I was just curious about I don't know if these are high priorities or low priorities but are you guys working on any other ecosystems outside of ethereum no but we have a very warm spot for Bitcoin and we're supportive of uh their research efforts um if if we uh saw or see a way to support putting Starks on bitcoin that would be uh you know something we'd

really uh try to help just because you know Bitcoin is uh the mother of all uh cryptocurrencies the first and it we are um you know on my Twitter handle I say I'm Stark Maxi so I'm that's that's what I'm maxing about I'm uh you know I really believe in our technology and its ability to help scale all blockchains and then go beyond blockchains into the conventional you know web tour web one or whatever web zero world so I'm very very um optimistic about that and if you're optimistic about this particular technology and its ability to scale blockchains you must be sort of supportive of all meaningful blockchains um as they're going to adopt it so I'm

curious um how would you rate kind of Cairo relative to other languages um and I mean we talk about you it evolving significantly over the years um there's a lot of developers it sounds like that are now onboarding recently into Cairo but um I've heard anecdotally it's still fairly complex and difficult to understand relative to other languages so I'm curious to get your perspective on that and how uh you can improve that okay um if what you're about is ease of expression of whatever because you want to write some app that I don't know what draws pictures on your computer you know you want something else but if you want um global scale for your application on

a blockchain it is not just the best it is the only um only path forward and I think this is what um identifies more than anything the community of developers they are there because they see the potential they see that this is a breakthrough technology made accessible in the fastest and most uh sensible way uh and they understand that being you know first steps um it's not going to be as well developed as other things I mean the analogy is suppose you're now you have your hands on the whatever the eniac machine and wherever it was and I don't know where it was first like one of

those very first machines or something so well if you want something really fun and and cool to work with that's not it but if you understand that there's immense potential to be Unleashed here you're just gonna jump up and down and get there now I do want to say that again and I'm saying this not as a developer but just someone who reads crypto Twitter the amount of enthusiasm and uh you know alignment that came after the Cairo 1.0 uh description saying that is very very much rust like and so on I mean there were multiple people saying you know this is going to be more pleasant than writing in solidity and you know didn't come from us but but so there's that as well

I I want to say that it's it's it's a very fast-moving uh uh uh Target um um I'm still stuck on the enlightenment I think earlier in addition to Innovations in cryptography can you say elated unrelated that's the next the next uh the next version of Miriam works but an appendix uh yeah I'm just stuck on that but I want to say this that the the the the the um the maturity of of the language Santiago is a very fast-moving Target and a strong function of of time so just in the anecdote back in April in Amsterdam there was a hackathon with 170 people

and um and there was one team that I met there and I met them a few weeks later at ECC that was starting at CC which was just like uh mine that was like like uh that was that was like uh anyway that was like a massive Festival of developers around Stark net and and uh um and one of the folks there that I met I had seen him only in like six weeks earlier and he said the development experience is just fundamentally different and I said what do you mean fundamentally different from a year ago he said no fundamentally different than it was at the hackathon six weeks ago so yeah a couple weeks ago this team building said we're developing ERC 1155 in Cairo and I said hold on guys I think there are

actually a couple of implementations already in place and you know and I send a message on the uh sort of the course uh the core devs for starknet on the telegram Channel and within two minutes there are three independent implementations including one from open Zeppelin but two independent ones as well so so uh uh yeah I'm I'm just saying it's it's moving very very fast right I have the you know if your listeners are gonna just take one thing out of it let it be this exploration sorry from the Oxford dictionary great happiness and exhilaration Elation what did I say Enlightenment that we can check out yeah I'll find your Enlightenment it's the

old uh Atomic Celtic version Middle English exactly exactly all right all right speaking of definitions I I learned about a new word this week which was uh layer three scaling I'm just starting to wrap my head around layer twos and here here we go with l3s and then I started reading about them this week and I as you guys are introducing the idea of l4s so now this stuff's just going over my head too much but can you guys get into uh can you guys get into l3s yeah L3 is is this is like in spinal tap you know where he shows the recordings it goes all the way to uh to 11. so so this is L3 and vitalik I'll published a post this week uh sort of expanding on on the notion of l3s that we published uh in December of

last year so we know we're very proud of the fact that that we innovate not only in sort of in cryptography and in the realm of zero knowledge but sort of in the broad sense and in blockchain research and L3 is one such example and the basic idea is a very simple one so instead of having these scaling Solutions settle their proofs uh by verifier smart contract on ethereum L1 do that onto a very fire smart contract that's deployed on a layer 2 system specifically on the public Stark net okay and by doing so you essentially move that layer whatever it is a stark X instance or a custom Stark net you move that one layer up now what do you get

for that what you get is even better scaling meaning even better gas per transaction and where it's sort of pretty astonishingly low numbers to begin with right immutable is minting hundreds of thousands of nfts server as well for a sub 10 gas per mint that's 20 000 X cheaper than they they were spending on ethereum L1 but you get even better scaling you get even better uh control over your costs as an application that spends uh ethereum L1 gas to sort of to settle its uh transactions and we think that most importantly you get better control and you get control that you don't have in a public network regardless of the question whether that network is ethereum L1 or the Public's Dark Net

applications it turns out enjoy living in a public sphere because of composability which is a hugely interesting capability and I think we're only scratching the surface of what can do that can do but there's a flip side to that point and that is what is often referred to as noisy neighbors right so if I'm I'm a game and something gas prices for me Spike because there's some action on uniswap or inversely if I'm uniswap and someone's minting Apes on the other side of the network and gas prices go nuts um you know that's your uh that's your dog right there so you know or a very big hamster L3 he heard about l3s he came running up

elated so I know I'm just elated that you can't hear Israeli drivers down in the street here honking like crazy at uh rush hour so so um so um the the the the thing that they get there uh by not operating in on a public chain as they get control they get control better control over their gas prices they get better control over their software the software version where the data availability solution that they uh introduce so imagine something like EIP 1559 okay this is like the canonical example that I always give right this was a very subtle and and uh and and uh thoughtful process that took a good while to introduce into ethereum and for good

reason right this is public infrastructure you want to make sure that you don't screw this up now you have to balance the the considerations and incentives of very different parties Etc now let's imagine that Yano has his Yano app chain and he knows for a fact that 1559 is a wonderful idea for him okay he introduces this into his environment into his chain you know just by by building consensus within his own organization he doesn't need to wait for the public chain to adopt this because he can do on his software stack as he wishes that level of control is something that organizations care for right the the blockchain space doesn't mean we're done with organizations as we know them today with business interests as we know them

today Etc right profits still matter and control still matters um and these app chains sometimes want a high wall around their Garden sometimes they want no wall around their Garden but they need that control and so we think that's what layer three will introduce so so I mean basically so you've got so the L1 like ethereum is for security and decentralization the L2 we'll use darknet in this example is for like scale scaling basically and like maybe composability and like general purpose scaling and then the L3 would be for uh just more customization like control and privacy and and and even more more skill and control exactly more scaling control you got it yeah that's 100 right hurry for the skeptic out

there how is this any different from Visa Network which is oh for all intensive purposes a big Consortium and oh there's a huge difference right there's the huge difference is that everything here you know it Turtles all the way down to ethereum L1 meaning if you say this is the logic that's going to be implemented on these smart contracts and that custom started in layer three they're available for all to observe this is the logic okay if that Visa decides maliciously to take all of Santiago's money and move it over to Yano it cannot unless Santiago signed all his assets over okay today with Visa right uh you know they can say well you know this happened you can call

the 1-800 number or you know uh take them to court uh but they uh uh you sort of assumed that they act in good faith here it's a very different equilibrium right even if they're 100 malicious there's no harm they can do other than censor you meaning other than decide that they don't want to serve you and then you opt out and go elsewhere right you go and transact elsewhere you try to bribe about something there you go minor or validator yeah yeah yeah bribe in the good sense of the word like just tip him um yeah yeah again I mean the thing um that we constantly hear and I'm not a skeptic of this but I do appreciate that some people look at this and say wait a minute like why do we need all this

stuff like you have hacks and you have other security kind of things that need to be figured out and other risks that are embedded into this equation um the idea though back the example visas I mean they they are sort of like you know they uphold their reputation in the business model that if they start doing things that you know it compromises their their integrity then they're sort of sabotage it's sort of like a fair Victory and that's something that I think a lot of people in crypto and blockchain don't consider often which is kick you out block you whatever but at the end of the day if they do that for a big subset of users then they're compromising their business model so I you know I I uh I guess some of the folks on this call are old enough and

bald enough to sort of recognize the subtleties of uh of the world as it is and and we're not of the camp that says you know sort of down with everything this is a new universe Etc the the world as it is exists today a has a whole bunch of uh benefits and advantages to it and B just won't just sort of vanish into the haze as I personally see it this that this uh set of of college trustless Technologies okay uh they shift the needle in a way okay and maybe in certain areas they have no value and and maybe in other areas they simply reduce the degree of of the the way in which the powers that be can abuse that power and maybe they diminish the

margins the premium that whatever insurance companies charge us uh uh I think it's a little too early to say how this will affect the world but I think it will overall be a very sort of positive force I want to say that um the the we live today in an age where the pendulum has swung too far to one side meaning um you know let's say 50 certainly 80 100 years ago most of Commerce was peer-to-peer and finances was very much peer-to-peer and um by now nearly all of it is electronic and centralized and that's not good for you know a variety of reasons and what blockchains are trying to do is move

this you know swing the pendulum back a little bit now even when it was peer-to-peer it's not that there weren't Banks and centralized institutions and they didn't was never fully completely only peer-to-peer but I think it was never um you know as centralized as it is today in the electronic age and if there are no blockchains it would be frighteningly so with with dire consequences all over the place yeah absolutely I really like the word you use there the a shift in the equilibrium um I'm curious in your discussions conversations with um perhaps projects or teams that are thinking about morphing their business model have you talked to a lot of

traditional Enterprises traditional companies that are interested in deploying software using it in some capacity we talk a lot about crypto native projects migrating to stockware and deploying there like maker and you know Ave but I'm curious if you have conversations and and the extent by which those are really serious of traditional companies looking to leverage this technology in some way shape or form we've had a gazillion of those um but first of all if we look at the you know empirically speaking uh you know if you look at our deployed systems or the things announced you won't see too many uh you'll see Zero you know S P 500 Blue Chips there and um

um even if you look you know a broader uh view at all of blockchains you know combined uh are there things beyond the POC level uh the answer is that right now not so much not yet um maybe you know nfts are slightly no no but I I I I I exclude like the the the uh the crypto Twitter announcements that I don't think that sort of yeah no but I'm saying there are you could say like in the nft world like you you do see some Blue Chip things whatever NBA answer or soccer it's it's a blue chip thing and it's making some appearance but uh and they're very deep reasons for that uh um that we can can go into but I think it's related to the previous uh

things certainly in the financial world so much is trust based and basically you're paying a very high premium premium for those you know big marble um you know stones at the doorway and um the the suits that you know are a social way of saying trust us um and that's okay it's reasonable but like uh trying to disrupt that um ecosystem with something that basically says no no you don't need to trust uh uh you know big suits uh there's something else it's just not gonna it's not that easy and I'm saying that now about all of crypto uh maybe we covered this but I want to go back to one idea obviously now there's a lot of

excitement around l2s a lot of them are up optimistic Roll-Ups we have optimism and arbitrum um arbitrary release Nitro that has some you know performance upgrades which are pretty interesting and you're seeing some really decent activity is it fair to say that these I mean if you talk to these teams I think they would probably tell you there is an upper limit there is an upper bound to how much you can scale these uh the ethereum blockchain through an L2 via optimistic roll-ups um but of course they would say hey well this is sort of like the transition of like in the similar to how like cellular networks evolve from like 1G 2G 3G 5G LTE is that a good analogy to then assume that's how blockchains will

evolve which is like maybe optimistic Roll-Ups are like 3G and then what you guys are doing is is LTE but it comes at a time where you know gas fees in ethereum L1 today are pretty low there's some activity um in l2s but you know you could I'm curious to I guess the question really is I'm curious to to get your appreciation for how um the sort of this this ecosystem will evolve in these sequential waves of like capex and and throughput uh do you feel that you're building something that you know there are some critics out there that say there's not a lot of usage and activity there's not a lot of really interesting applications being built yet like why build LTE when you don't even

have like a really nice smartphone yet I think um our role in what we need to do is and I'm saying now this about you know as speaking as one of the l2s but but I think it's the same I would say that it's really the same for our colleagues they're also are our competitors but they are our colleagues and part of the same ecosystem at optimism and arbitrum and other places um each should work as hard and as fast as possible to offer its scaling solution and then you know markets in the world will decide and that's fine like and we'll accept whatever judgment the world passes and then you know each team is going to need to update or decide what it wants to do next let me

give you know you mentioned one example and I think at some point in the past the various folks referred to it and said you know maybe optimistic Roll-Ups are something for the midterm and the long term solution is going to be Stark net and you know things like that and yeah that could be could be that like uh just like we see today there's a lot of uh um a lot of teams are now taking their solidity code and deploying it on polygon uh so you know maybe after polygon they go and deploy it on on optimism and arbitrum and and this works and then maybe at some point they in our case they're gonna you know for better or worse they need to rewrite it to to make it more scalable they Port it over to Stark net that's certainly one option that it sort of is the in the in these

phases uh there's another uh possibility which is that that we outpace the other things and you know once uh we have the performance and we do the Regenesis it'll actually see us eclipsing the others that's another thing a third one that that I hope for us doesn't happen is that you know we never uh lift off and like everything is everyone's happy with whatever it is polygon or uh or uh optimism or we our role as you know leaders of one or coming from one L2 is to just make sure that this is the best technology and most ripe and usable by developers um and then we're gonna let the world uh pass judgment on how it wants to use it this has been great guys

um I think we should wrap it here this seems like a good place to wrap I have one closing question that I'm going to ask both you guys and I want it's a it's a prediction question um Yuri I'm going to start with you here in five years what percentage of crypto is built on there's three buckets here what percentage of crypto is built on l2s verse L ones like the Solana and Avalanche Etc l2s could be you guys or competitors and then the third Buck would be app chains like folks who go over to Cosmos oh boy um I have no idea I have no idea I think that the world as value accumulates in these applications is going to come to

appreciate perhaps in painful ways the value of secure systems and so understanding that a cosmos chain like say luna and then put aside the the the shenanigans and the criminal activity but a network that's secured by the price of an asset that can suddenly tank is a very dangerous thing so I think that that uh ethereum dominance in this regard I think is going to increase over time uh compared to the cosmos chains but there may be some Innovation there where they demonstrate the ability to bestow security from one of those that have changed on to others and I think there are expert there are ideas and experimentation in that regard and you

know I hopefully further explore that um but for now I think it's going to be uh an ethereum dominated world and I think that layer threes are in fact going to be uh that sweet spot where applications can balance their needs to customize with uh with their desire to scale in a in a verifiable and transparent fashion Eli I'll throw the same question to you I have no idea about uh yeah I think that in five years people will be speaking in other terms I don't know what they'll be because but but uh but I I will make a prediction um I think that in five years um our core technology

um Starks Baby by then even ZK Starks uh by then maybe we'll actually add privacy in the world is going to demand it and uh you know things like Cairo 2.0 or 3.0 or whatever is you know that language is going to be called like the whole tool chain that uses um Starks for exponentially increasing um you know scale and offering Integrity um that I'm very uh confident of you you'll see that uh 90 of uh the way or you know the amount of transactions that transpire on on all major blockchains and maybe even it's going to start

percolating outside um to these permissionless or decentralized ecosystems ninety percent of uh the actual game is going to be um proved by uh Starks or ZK Stark's proofs and probably coded in some Stark friendly language which could be again Cairo 2 or 3.0 depending uh or some other uh dedicated language that is very uh that compiles into really good um an efficient start proofs yeah and you can you know you can remind me in five years I think that's where we'll be yeah I guess maybe I'll throw one last question

um which is I think Stark where like I think like zero knowledge proofs have been in the conversation for such a long time they're perhaps one of the more exciting um elements of cryptography at least is my appreciation um what are some of the things that you you've been doing this for for a number of years um what are some of the things that you thought would probably move faster if you were to say like look back over the last five years you would have thought gosh I thought that we would have developed X or Y much faster and then conversely what are the things that surprise you that have developed much faster than perhaps you would have initially thought and I asked this because I think like sometimes we're really impatient we get excited about the technology but we're impatient or we sort of it's that like

you know we underestimate what can happen in 10 years and overestimate what can happen into kind of like Mantra that Bill Gates introduced I'll give him my answer so in May 2013 um you know no one was mentioning scaling through proofs I'm proud to say I mean I was the first one to suggest it and the you know I challenged the world to find the place where this was said earlier and I said it at the Bitcoin conference in San Jose and that's where I was red pilled and in that you know half hour talk that is recorded and you know I'm much younger there um I said um by the end of the summer by the end

of the summer um and this was 2013 on May 2013. I said by the end of the summer we'll have basically the code for it wasn't called um uh starts back then but what essentially became you know later on called Stark's I said by the end of the summer um and then it took uh five years to release the relevant academic code in the Amor years this and that so like the speed at which uh code is built is I've been you know constantly over optimistic on this but where uh I'm completely blown away is recently like with the amount and speed of uh adoption of uh starknet and Cairo and developers flocking to it I and I

think others I'm speaking for myself I I didn't accept expect this much adoption rate so quickly so soon awesome great guys Santi anything else no unless all right you have a response to that uh or or just uh something in your mind yeah I'm I'm not as optimistic as Ellie so so like the the super macro view when we were undergrads that was a fair number of years ago uh zero knowledge proofs were with pride taught to us as something that is completely theoretical and could never ever be computed the fact that there is now a general purpose programming language that allows you to to prove any computational statement that this is in production that this has settled hundreds of billions of dollars

over the past year on our software stack that there are hundreds of people writing higher programs that we don't know and we never taught Cairo to uh mind-blowing to me so um I'm happy as a clown elated as as developers yeah I mean yeah elated elated are there are there any other interesting maybe not even in crypto like outside of zero knowledge proof so are there any really interestingly interesting things maybe not even happening in crypto yeah yeah I'm I'm working on on homeless popsicles and that's the technology that I've been perfecting for a good number of years and you know you're I thought we were going to invite you back on the podcast

but no more my friend that's it you're trolling keep saying it's the last question then you keep asking questions all right we're wrapping it guys congratulations on all the progress um yeah we really appreciate you coming on in here and and thank you very much thank you gentlemen amazing thank you very uh intelligent uh interview and very um questions yeah yeah yeah this was wonderful guys happy to chat anytime you want yeah we would love to have you on I think uh the pace by which you guys are shipping and developing it seems to be much faster now Cairo and really excited to see where that goes especially how easy it is now it sounds like for a lot of developers yeah sure

what are you doing next week same time we'll have a we'll have a ZK regular stand-up uh but uh okay thanks for coming on yeah I really appreciate it awesome guys well enjoy it thank you guys goodbye [Applause] [Music]