Join Delphi Research today and immediately get access to our full Member Portal!
Join Delphi Research today and immediately get access to our full Member Portal!

Transacting on StarkWare’s Live ZK-Rollups Is Orders of Magnitude Cheaper With the Security of Ethereum, StarkWare-Powered Gaming, and Building a Vibrant Ecosystem of dApps

Oct 28, 2021 · 58 min media

By Tom Shaughnessy, and Medio Demarco


The Delphi Podcast Host and GP of Delphi Ventures Tom Shaughnessy and Delphi’s Head of Research, Medio Demarco, are joined by Eli Ben-Sasson, and Uri Kolodny, Co-Founders of StarkWare, a suite of validity-based blockchain scaling and privacy solutions. Topics discussed include how StarkWare stacks up against other scaling solutions, achieving low transaction costs, developing a vibrant StarkNet dApp ecosystem, and much more.

Social links: 



Interview Transcript:

Tom (00:17):

Everyone. Welcome back to the Delphi Podcast. I’m one of your co-hosts and I help on Delphi Ventures, Tom Shaughnessy. I’m here with my co-founder and partner Medio DeMarco, who is also our head of research. We’re thrilled to house Uri and Eli from Starkware. I’ll let Medio give a brief introduction of himself and then we’ll go to Uri and then Eli.

Medio (02:19):

Yeah, thanks Tom. This is a podcast I’ve been looking forward to for a while now, a quick recap on myself. Come from a traditional finance background, having worked at Bloomberg and Deutsche Bank, caught the crypto bug back in 2017 and then left my old banking job in 2018 to co-found Delphi with my partners. And it’s been a wild ride ever since.

Tom (02:40):

Love it here. Uri you’re up.

Uri (02:44):

So I’ve been an entrepreneur for the past 20 some years, I’ve known Eli for longer than most of you guys have been around. And late 2017 at least said, “Let’s do this.” And so here we are.

Tom (03:00):

I love that, Eli. What about yourself?

Eli (03:02):

Okay. So I’ve been doing research on the kind of technology that we ended up using since roughly 2001, first within academic circles, and then started moving more towards blockchain. I caught the blockchain bug in 2013. I was a founding scientist of Zcash around 2015. And this technology Starks that we’re using and Starkware we developed it shortly after. And that’s the basis for our company.

Tom (03:38):

I like that. It would be helpful for the listeners who aren’t up to speed on Starkware I guess, for a quick or a brief introduction before we dig into the nitty gritty, not sure who wants to take that one, but we’d love a quick elevator pitch to get people up to speed about.

Eli (03:52):

I can take that. So you can think of the blockchain as this computer that everyone wants to use, and it runs at a pretty slow speed. So like think of the early computers of the 80s or something, everyone wants to use it. So it gets congested and getting it to run your transaction is very costly. 

What we do is speed up this computer that’s when you’re thinking of it, and the way we speed it up is that, instead of asking it to run your transactions, mint your favorite NFT, do your latest trade on some automatic market maker, instead of asking this very slow computer to do that, we are asking the slow computer to verify one of our proofs, one of our stark proofs. 

And the magic of the technology is that this proof is exponentially smaller than the amount of computation that it asserts to. So basically we now have this very slow computer that everyone loves, verifying a proof that it tests to tens of thousands or hundreds of thousands of transactions with the same level of security. And that’s how we speed up the blockchain.

Tom (05:03):

It’s a great overview. It sounds so simple. Like you make it sound like it makes so much sense, but the work and the math and the developer activity that has to go into making something like this work is extreme. Can you guys give us a brief introduction on how long you’ve been working at this and just how complex the math and decode is behind this?

Eli (05:25):

Okay. I’ll take this one. So we’ve been at it for four years. We’re now a team of nearing 60 and very brilliant and hardworking folks that are in the combination or have the unique combination of being very strong on the math and cryptography side, but also being very, very good engineers and researchers and product managers and things like that. 

So four years, a team that grew up to 60, there’s a very, very deep technological stack behind it, but part of the achievement that our team has reached is that, all of that is now sort of abstracted away through an interface that’s very easy to use and program in and deploy on. So yeah, there’s this magic-

Uri (06:19):

I want to quote this out on the starting at discord yesterday, who said that verifying the CAPTCHA is harder than picking up Cairo. A, that’s a cute quote, but B, it also goes to show… we’ve put an awful lot of effort into trying and wrapping this very sophisticated, very cutting edge math and cryptography into tools that any developer can use.

Tom (06:51):

So when a developer’s working and they want to build on Starkware, do they have to know both Cairo and East language? Or can they use one or the other? I’m just trying to get a sense of the hurdles there. Because if Cairo is easy to use, that’s fantastic.

Uri (07:04):

So Cairo over the course of time is becoming increasingly more powerful in the higher level and easier to use. So that’s one thing we’re hearing over and over again from developers. Another mind folks have put together a very powerful team called Warp and they’re developing a transfiler from solidity to Cairo. This team led by Greg Vardy. It already supports general computation, within a couple of weeks time, we expect them to announce that it supports composability. 

So now developers have both these tools in their hands and they can pick and choose what they want to use, fundamentally and probably in the longterm. Writing native Cairo code will result in more performance code. One comment that’s worth emphasizing in this context has to do with the pre-existing solidity code. That one would consider porting over to Layer 2.

And the important thing to understand there is that, the interesting stuff and the important stuff has nothing to do with syntax, solidity syntax versus Cairo syntax. The main point there is that those solidity contracts on Layer-One were written under a very particular set of computational constraints, and that is Ethereum’s gas limit per block. And so people devise all sorts of very clever and very sophisticated schemes to live under this extremely low glass ceiling. And built AMMs and took great pride in the fact that between version two and version three they’ve cut gas costs by 20 or 30%, et cetera.

Now there is suddenly offered to move to Layer 2. Now, moving over this open field computationally speaking, into the screen, open, vast open field, moving over with the same set of constraints that you’ve lived under on L1 makes very little sense. And that’s through the prime example of the folks who realize that early on and acted on that realization as the dydx. They did not repeat, did not take their Layer-One logic and just translate it to Cairo. Instead, they started from a blank sheet of paper, and they said, “What is that application we would have built, starting a fresh?”

And that is what they built in. And the result is not only a far more skills system, we’re talking there about 600 gas per transaction on StarkEx, they were at two to 300,000 gas per transaction on their Layer-One. So the result is not only a fundamentally more scalable system, it’s a far more sophisticated and capital efficient financial tool.

Medio (09:53):

You brought up a few good points there that I’d like to dig into a bit more because they’re really big differentiators for what a Starkware L2 can offer versus the field. One of them being the complexity of the computation that’s involved, one sector that this has clear utility for is gaming. Could you maybe just speak to how new game studios and existing ones are working to deploy game logic on top of starting it?

Eli (10:22):

Yeah, there are a number of projects that have already announced that they’re doing this. Influence is one of them just yesterday. 

Uri (10:33):

There’s a podcast going on right now. 

Tom (10:40):

[crosstalk 00:10:40] podcast.

Uri (10:40):

Oh, exactly. Hosting Louie Goodman. Our ecosystem lead to talk about.

Eli (10:47):

As we’re recording this.

Uri (10:48):

As we’re recording this. 

Eli (10:51):

And what’s nice about gaming in StarkNet is that, when it comes to gaming, you really want a little bit more computation to be done during the gameplay. And as we said earlier on Ethereum and blockchains are like this very slow computer. And the one thing that StarkNet is doing far better is speeding up computation. So we’re going to see a lot more of that and a lot more games deploying.

Medio (11:21):

That’s great. And the other point you mentioned Uri was just the cost of transacting. And I think one of the secret sauce, if you will, to what makes Starkware unique is that, it kind of benefits from economies of scale where the more transactions that are occurring, the cheaper it gets for everybody. Could you maybe speak to that point? Because, I know that is very unique relative to other Layer-Ones or a Layer 2 is where, as more transactions flood in everything kind of worsens, whereas that’s necessarily the case here. 

Uri (11:54):

Yeah. So, first of all, right off the bat, just to quote some numbers, one I’m quoting here, this is not crypto Twitter. This isn’t a spreadsheet. This is reporting from the field live from Ethereum [inaudible 00:12:08]. So, rare and Immutable  are minting batches of 600,000 NFTs in for lidium data availability mode, meaning with off-chain data at less than 10 of gas per minute. Okay. Now this is 20,000 times cheaper than what they used to do on [inaudible 00:12:28]. So this is sort of like right off the bat, the basic functionality.

Now what we are seeing in the Medio, exactly to your point, there are fundamentally economies of scale that fall directly from the basic capabilities of the tooling that we’ve developed in. First and foremost, here is Cairo our train complete programming language.

And this allows for a service we call SHARP, which is shared prover and SHARP is now in production serving, the same instance is now serving Immutable  and Servair and DeversiFi and StarkNet at itself. And we hope to bring dydx onboard soon. We intend to actually put forward a proposal to the dydx community to motivate this. SHARP basically says the following, because you end up essentially, you’re a spec because it’s logarithmically cheaper to exponentially cheaper to verify a proof compared to the competition that you’re proving off chain. You have a very strong incentive to create very sizeable proofs for huge computations.

So your proof, instead of a testing for Tom’s application, with his transactions and for Eli’s applications with his transaction, what you could do is in fact, create a single proof that attests to both, or to any number of applications, this for an applications you have no need to synchronize amongst them. They have completely independent logic, et cetera. Now, all these applications were bundled into a single proof, and this one proof is now verified in one go on chain. And so Immutable, Sorare, DeversiFi today are splitting their gas costs on Layer-One. What does this mean? 

This means that if video is coming in with his new game, and you’d have like 20 users of date, okay. From day one from the very first proof that is generated to scale his application, he’s already writing on the marginal cost of the entire Cairo ecosystem. And that’s a very powerful network effect.

Medio (14:43):

Absolutely. Funny enough when we started Delphi back in 2018, one of the first reports we did was on zero knowledge proofs and Starkware was one of the projects we focused on. And one of the aspects we covered in there was just that, we don’t have to get into the deep technicals of everything and all the different flavors out there that zero knowledge proofs manifest themselves in. But a Stark proof is the biggest one of the field. 

And, that being used to batch a few transactions, it doesn’t make much sense, but to your point about the proof size growing logarithmically as you pack in more and more transactions yeah, when you’re batching and trying to compress an entire off-chain ecosystem, that’s where it really shines. It’s a really powerful thing to offer because again, it is unique.

Uri (15:37):

And users at the end of the day, the only thing they care about, the only thing they should care about is the amortized gas per transaction, because that’s what they’re going to incur on a personal level.

Medio (15:49):


Tom (15:50):

To Medio’s point there like, what is the implementation speed for onboarding the apps to users, like getting a dydx involved. You said that they kind of went back and wrote their new code. What exactly does that look like time-wise?

Uri (16:02):

So let’s pick on Sorare, for example. Sorare went from, we went from a handshake to Ethereum Mainnet in less than two months, one and a half engineers from Sorare. Now we keep saying this to peer at Sorare that that’s one very powerful engineer, but that’s one engineer. So that’s in terms of StarkEx, StarkEx now has an open access playground. Anyone can interact with extensive documentation. And so onboarding now is I think is a very straightforward thing. On the stark net front. We’re seeing things that frankly blow our minds. 

And I’ll say, I keep tweeting about it. The most exciting thing for me is that people are doing things that we never imagined. And there, the permissionless of it is just shines and all its beauty. So, this very talented guy with a brilliant handle and guilty gills started proving all sorts of physics simulations over StarkNet. And I think literally a couple of weeks ago, I don’t think he knew Cairo. Now Sean Han tweeted out today that is StarkNet JS framework now has the 100 downloads and three contributors, and he didn’t know Cairo two weeks ago. So the whole thing is happening super fast. It’s literally unfolding in front of our eyes right now.

Medio (17:35):

Those tweets from Guilty use, those are really interesting. They came up on my feed as well. Truly just fascinating stuff. And the speed at which these people are learning Cairo, starting to deploy, I think that’s something that’s really surprised myself over this past year. If you rewind to Q4 of last year, you had optimistic roll-ups in one corner, you had ZK-Rollups in another corner and both had their strengths and weaknesses. You can make a good argument that ZK-Rollups were a better long-term solution for a variety of reasons, but near term might have some trouble getting traction, just they were expected to be more difficult to build on and migrate over from Ethereum Layer-One.

But fast forward to where we are at right now, the optimistic roll-ups hadn’t really… they were supposed to have an easier path to gaining adoption building on, and they hadn’t really run away in that regard either. While at the same time, Starkware already has. You look at dydx and the transaction volume they’re processing. So moving forward, I’d be curious to get your thoughts, how do you see what you’re building stacking up, not just to optimistic roles, but also other Layer-Ones as well that could offer, strong scalability and benefits as such?

Eli (19:02):

I want to say, first of all, I have a lot of respect for the optimistic roll-ups for optimism and arbitrary. I think they’re doing a terrific job, but the inherent theoretical limitations of the optimistic role of technology are such that, it reaches lower scale than what you can get with a ZK-Rollup. I’m sorry, but that’s the math. And what we see playing out is exactly this thing. So, I’m eager to see, for instance Reddit rolling out on an arbitrary, and I just want to see what kind of skill they’ll be able to do.

I think it’s going to be a tremendous challenge to reach the scale that Reddit needs with the kind of technology that is optimistic roll-ups whereas, I’m very confident with the ZK-Rollups they can do far greater scale. Again, it’s not just because we have to, ZK-Rollups it’s rather because we think it’s a better technology that we’re using it. Now to your question of just looking ahead and where we’re heading at. We know that this technology, because it delivers integrity, which is vastly needed in our world, and it delivers it at tremendous scale.

And it can also give you privacy through the ZK aspect, is something that the world needs. And it needs it not just in blockchains, it needs it all around. It needs it within the context of citizens versus their governments. And with respect to the way we deal with our own data, we need to be the source of truth. We need to own it back again and not have these big data guzzling corporations that are holding all of our data and profiting from it.

And it’s clear in my mind that ZK is starks. And the way we made it accessible for developers and for the world is going to play a pivotal role in this transformation of basically giving back ownership of your digital life to yourself, to the citizens out there. So it’s not going to stop. It starts in L2, it’s going to be a technology that’s adopted farther and wider within the context of, for instance, CBDCs. And I think at some point there’ll be calls from governments and societies to use this kind of technology in order to empower citizens and society and decentralize a lot of the way we interact with healthcare, COVID contact tracing, all sorts of things that are outside of blockchain.

Medio (21:55):

Absolutely. And I think that’s an important point to stress, the scalability that Starkware offers, it’s more than just DeFi, NFTs. there’s so many applications that haven’t even emerged yet that can benefit from it. One, when you mentioned the word integrity, one synonym that came to mind was validity. And I think that’s also something that is a strength of what you guys have built. You use validity proofs, right? When that proof gets posted on the L1 and it’s good to go right away. There is no you know game period where you have to challenge potential frauds, it’s correct. 

And that is like the magic math behind how ZKPs work. And it has its benefits. If you’re on a ZK-Rollup on Starkware, there isn’t that seven day wait period that system’s built using fraud proofs kind of have to deal with. And from a UX perspective, it can be really smooth.

Uri (22:58):

By the way, it’s not only UX. It’s interesting to see how the whole blockchain is facing it. I understand the motivation it’s early on, it’s a nascent ecosystem, but all sorts of considerations that, in the real world on Wall Street would have been like an immediate end of discussion, in the context of blockchain discussions, we’re sort of kicking the kind of thing. Well, now the one week withdrawal period and the capital efficiency that implications of that delay, we’re just going to ignore that for now. In steady state, it cannot be ignored, meaning someone has to pay for that. And it ends up in some way or another being the end user.

Tom (23:49):

I think the community building aspect, you guys mentioned is important. Like, Uri, you mentioned earlier that you’re seeing kind of unforeseen use cases, but it’s not just relying on you guys building this out. For disclosure, we’re investors in Immutable  and dydx, but Immutable  is built on Starkware and they have gods on chain, they’ve built a guardians. And they’re essentially operating, they essentially built an operator layer on top, which allows developers core access that makes it much easier to onboard. How do you think about other platforms or projects building on you, which has greatly kind of increases your kind of target market here?

Uri (24:25):

I think that’s a super important thing in terms of making this technology broadly used and share the story, the experience we had with TikTok talk a few weeks ago, they reached out and they said, “We want to mentor our first NFTs. And well we have 12 days to go.” Was what they said. And we put them in touch with the remarkable team at Immutable  and they got working. And the first TikTok NFT was auctioned off for about $100,000 a couple of days ago. So, Immutable , I think is one example of a very powerful extension of the capabilities and value of this software stark. 

I should also mention in this context, not only the people building on top of our software stark, but in fact, building with us, the software stark, and specifically in the context of StarkNet this is an ecosystem. And so there are two full node efforts going on for StarkNet full nodes. One is the powerful Aragon team, [inaudible 00:25:37] building fermion, and there’s the equilibrium team building, another full node. We mentioned the warp team, another mind are also developing a block Explorer for StarknNet. There are multiple teams ramping up their code auditing capabilities to be able to audit Cairo code, the figment team were going to be providing API services. So, there’s a whole eco, it does indeed take a village and we’re happy to be part of that village.

Tom (26:09):

That’s phenomenal. Is it kind of surreal for you guys to see TikTok, like one of the most addicting and well-known platforms using you guys? Like, did you envision that when you guys got started to have this level of uptake right now?

Eli (26:22):

No, but it’s concretely very helpful because my kids finally, they don’t understand what we’re doing, but they find that they have some respect for it because TikTok is using it. 

Tom (26:35):

Yeah, no, it’s interesting to see. Just zooming out on the adoption spectrum. What would give you guys a level of accomplishment one year from today looking at StarkNet? Like what do you want to see ported over? What do you want to see built? Or is it more of a question of, let’s find out?

Eli (26:55):

I think first of all, it’s definitely let’s find out. I think that the things that a year from now we’ll see as being very successful on StarkNet are things that we can’t predict, predict, and that’s the fun of it. The one metric we’re most interested is basically volume of computation. We’re already tracking it, the amount of computation that our systems can support, the latency, the amount of usage of proving. That’s really our core metric that we’re tracking.

We’re already very happy to see. Our effective TPS as we’re speaking now is far higher than Bitcoin and Ethereum right now. And it’s taking just a very tiny fraction of Etherium’s bandwidth in terms of gas. And it’s going to grow by a few orders of magnitude. So this is the thing we’re looking at and tracking.

Medio (27:51):

Absolutely. And there’s a thing on the horizon, that’s going to make it even cheaper outside of what you guys, yourselves are building. And that comes down to the data itself hosting it off chain, posting it on chain, with the transition from the current state of Ethereum to ETH 2.0, one aspect of that is sharding the blockchain, I believe first up is sharding just data storage the most basic sense, which is a great benefit to you guys. Like you’re posting big proofs that are compressing, tens of thousands, hundreds of thousands of transactions. And now there’s more on chain space to post that too.

How are you guys thinking about data availability, sharding, and also, this might be a good time to transition into the difference between [inaudible 00:28:44] rollup and Ethereum?

Uri (28:46):

Sure. So, let me say a few words about data availability. So we really think, and we described it in those terms over a year ago. We think of it as a very rich design space, sort of a spectrum. So on the one end is roll-ups with data being on chain and that’s of course the most secure, that’s also the most expensive. So those are roll ups and stark operates in roll mode. Dydx is using StarkEx and roll up code, that’s their choice. 

Uri (29:16):

StarkEx also comes in Validium mode where data is stored off chain and a set of validators signed per batch, that they have the data off chain, that maps to the current commitment of the new state of the system. And absent those signatures that batch cannot be set as a valid on chain. Okay. So that is Validium. What was interesting to see with folks who are using Validium and these are at the moment are Sorare and Immutable and DeversiFi is that they, I think were deliberating up to the very last moment, whether they want to run in Validium mode or in roll up mode. 

And to us, that’s an indication that there’s… aside from the Immutable sort of the pontifications on crypto Twitter on what is the true cause and what isn’t the true cause and all that stuff, there’s the real world, the real considerations and users and costs and those trade offs, and those deliberations are real. Now, what we’ve designed in the process is something we call volition. And volition is very interesting because it basically moves the decision from the application to the user at the single transaction level.

And this is coming on to Mainnet in less than four weeks with Sorare and Immutable . And so with volition users can choose for example, if you have some very valuable NFT and you want that on chain, it’s more secure, et cetera, you’re willing to incur the transaction costs, et cetera. That’s your decision. You have a bunch of other NFTs that are not worth as much. You want to keep those off chain, that’s perfectly fine. Something appreciated and value and now you want to move it on chain. No problem. 

I want to stress that this is not only for NFTs, right? Then let me describe another use case from the fungible, from the defined space. You’re a high frequency trading firm. You come into the office on a Monday, you want to trade a ton of times throughout the day. You don’t want to incur those gas costs. So you can do this in Validium mode, or off-chain at the end of the day, you put all your data on chain and you’re good to go.

Medio (31:37):

And that is a point worth re-emphasizing. It’s not the applications that are choosing the security of the mode they’re operating in, that is being given to the user themselves. And that’s really powerful.

Uri (31:50):

Exactly. And we will see very soon, whether folks are, what are they willing to pay at what price points? We’ll have a very beautiful experiment sort of unfold.

Medio (32:06):

And this is a good point to ask a question that I’ve been dying to get an answer to. Even though I’ve seen some potential answers in the Twitter comments. So the difference between Validium versus as ZK-Rollup, ZK-Rollup more secure all around, and Validium less so, but the trade off is that Validium should be cheaper. Is it correct that recently the ZK-Rollup transactions were actually cheaper on a gas basis individually than the Validium ones? And if so, how is that possible?

Eli (32:38):

Right. So I actually posted a riddle about that. How could it be that in roll-up mode there was a lower amortized gas costs and for the Validium? And the answer to that riddle is, so the theory is still holds. If you take the same system in the same pattern of use, and the same size of proofs, Validium will be cheaper amortized than rollup. Okay. But what happened there is that the dydx system is the one running in roll-up mode. We are using for it much larger batch sizes. So the amortized gas cost goes down as the batch size grows, that’s factor number one.

And the second thing is that when you use a validity roll-up as opposed for instance, an optimistic roll-up, the data that you’re rolling up is only the final state diff. Now, what does it mean if you have a small number of accounts that are doing all of the trading, then it could be that the state diff, which is the part that goes in a roll up on the chain, whereas in Validium it doesn’t go on the chain. That part becomes very small. Let’s take it to the extreme example. Each batch in dydx is 12,000 transactions.

If all of those transactions were touching just two traders, even though it’s a roll-up, the state diff would be very shortly because it would talk only about two sales. Again, this is something that in an optimistic roll up, you would still need like 12,000 signatures and witness data. So those two factors combined, the fact that there were very few accounts that did a lot of the information, and that the proof sizes in the roll-up mode were just happened to be larger. That’s how you got to that anomaly that advertised gas costs in roll-up was lower than in Validium.

Medio (34:40):

That’s great to know. When I first seen that tweeted out, I was impressed, but also befuddled as like, how is this possible? But your point and I guess to phrase it a different way when you’re saying you’re just posting the state difference, you’re essentially just netting it all out and just posting the net difference there. Instead of, taking 10,000 transactions and all those signatures, all that data, you’re just doing this, the net between these two here you go.

Eli (35:07):


Tom (35:11):

Okay. Switching gears a little bit. You guys are obviously in the limelight as one of the most critical and important piece of the technology building on Ethereum right now. So of course, you’re going to attract a good class of smart critics. What do you think is the most legitimate critique of StarkEx or Starkware? And I guess, what would your response to that be?

Uri (35:32):

I think the most important critique has to do with the nature of cryptography, which at the heart of it is, it always sort of puzzles me and amazes me that at the heart of it’s faith based. So the strength of the hardness of hashes is something we believe in and we have sort of cumulative experience suggesting that it is safe and secure. We don’t know for a fact. And I think that the most succinct and intelligent presentation of that case was made by David DvorakForic, when he presented in surplus sessions a couple of years ago, when he said, “Guys, this is all very new stuff. All these zero-knowledge proofs.” And in sort of a historical perspective, that is true. It is new stuff.

And in that regard, the assets we have locked almost billion dollars that we have locked up in StarkEx is an open bounty program, just like any other smart contract. And so all these things, people’s confidence will increase over time. What’s important to appreciate in the context of our cryptography, is that in areas as a far better and more eloquent explainer of that is that we rely in fact on far leaner and far more battle-hardened assumptions than the other zero-knowledge proof systems that our folks are trying to develop these things. Eli, do you want to sort of-

Eli (37:04):

I want to also try to answer this question, I want to say yes, there’s a lot of critique and good criticisms. And we like that. And we take them very seriously and we try to address them. So, one thing that was said about our technology is first of all, you need to be Starkware in order to build or understand this technology. And we listened to that. So, Cairo, StarkNet, they are answers to this very valid criticism. 

Another thing that was leveled at us again, has a lot of strength is that like only Starkware is like this a proprietary corporation and, how are the masses going to use it? And again, StarkNet is exactly a reaction to this and just to make it very clear, StarkNet will be decentralized and anyone will be able to run proofers and sequencers later, like let’s say within nominally one year from today, or even less. And this is again, responding to this very valid criticism of we can’t have it, that you guys are the only ones running it.

And of course there’s the very valid criticism as to how do you know that the stark is sound from a cryptographic and code wise point of view? And that’s a very valid concern. So in addition to the more standard things such as doing code audits and training people to be auditors for the new programming language of Cairo, which is something that we’re proudly doing with several teams, we did another thing that, I hope other offers of cryptographic protocols would do, which is, we went through a very serious auditing of the math. 

We proved via proof assistance that Cairo and its set of constraints is actually sound. That’s actually a topic that, there’s a white paper on that, and we’ll discuss it there in our next stark at home, which we’re right now advertising. So we’re going these extra steps to address also the validity and soundness of our tech stark.

Medio (39:21):

That’s great color. And this is a conversation I’ve had as well talking about the Starkware and all the different pieces, the proofer component in its current state. I believe there is just one entity generating the proof and posting that on chain, in the future that’s going to move towards decentralization, which will be great. And then, that’s a point that gets brought up by the other side. And my response is always, yes, that part isn’t in the ideal state yet, it’s going to get better, but at the same time, if you look at all the other benefits you get, it’s still a net positive all around right there. 

It’s not custodial. It’s got the base layer of security. It’s orders of magnitude more scalable. These attributes make the current trade-offs that had been made and very reasonably, so worth it. And then, yeah, the shift towards decentralizing StarkNet and all its different forms will be something I’ll eagerly watch on that front.

Eli (40:18):

Yeah. Regarding the approvers. I just want to say we open sourced already one prover within each stark, we strongly support all of the other efforts that others are doing, working towards a writing starks, there’s the Winterfeld project under Facebook Navi led by Bobin, the mirror protocol folks are also working on various components. They just published a beautiful implementation of a very new result that we have with ECF 50 elliptic curve at 50. The Hermes team are building and starting to work on things that are related to starks like fry and so on. We strongly support that. Then we will also release a more coded proofers for folks to use.

Tom (41:03):

When do you guys think, building on Medio’s point there about, and your guys’ points on how battle-tested this will be eventually people will get to a point where the developers do you guys as AWS, right. They come to use your product suite, they build out an application and it’s built on Starkware, so nobody really thinks about it. It’s just the industry standard. When do you think you guys will get to a point where people don’t really care about the code anymore? They don’t care about the math. They’re just here to build.

Eli (41:29):

I prefer Ethereum by the way to AWS. 

Eli (41:32):

So just like, Ethereum, how do you know that [inaudible 00:41:34] and EVM and everything is sound right? At some point, you reach a level of comfort [crosstalk 00:41:42] pretty much there. I think we’re pretty much there. We’re not being asked too many questions these days about the soundness of the core technology, but I don’t know, it’s going to take a bit more time maybe. Every year that passes and more volume accrues and there’s no fundamental bugger problem, I guess sort of bolsters this position.

Medio (42:04):

Absolutely. And we’re talking about for the complains here, and of course in crypto, we’ve experienced so many delays, so many setbacks in terms of scaling plans specifically. But one thing that stood out to me when I was reading through recent posts from you guys is that, certain aspects have been ahead of schedule and that’s unique. One such thing was composability. I know that was a year ago, levied against stark. Whereas, the criticism that composability within the ecosystem would be a challenge. And here we are, that should be coming very soon. And it’s already arrived. Could you maybe speak to that point?

Uri (42:43):

Yeah. We announced the stark roadmap in January by June, we were already on a public Testnet that with supporting general computation in a permissionless manner. StarkNet on Testnet today supports composability, has on chain data there when they are two messaging. Basically has all the components of a permissionless network and that’s the stark in an alpha that is coming to a Mainnet in a few weeks time. Indeed months ahead of time. I think this goes to a couple of things. A, we’ve spent a good number of years developing a very powerful toolbox and very talented folks have been developing this toolbox. 

And so now that it’s available to them, they can move very fast. And the other thing is there is a culture at Starkware that we’re proud of. We take our deadlines very seriously. And so just in terms of SK-Rollups being with a permissionless public test that for months supporting general computation, I believe there’s no one else who falls under that category and that’s something we take great pride in.

Tom (43:58):

One question I had for you guys was in building a community, it’s a little bit harder when you don’t have a token, because it’s harder for people to conceptualize. For me, if I’m building ETH, I have 100, I have a clear link and economic link, for you guys it’s a little bit different because projects can build on your tech and be successful through their own token. And that’s their link through, on being on fantastic technology. But there is a little bit of a gap there, I think, or maybe I’m just viewing it under a retail or an investor lens. And I shouldn’t be, what’s your take on building a community without having a token?

Eli (44:33):

On token, I’d say no comment on building a community without a token. StarkNet is going to launch in a month, there’s no token on it in a month. We’re optimistic about it’s adoption based on what we’re seeing now, but let’s talk again and let’s say three months, four months and see where we are at.

Medio (44:55):

Say no more. So one question I’ve had, looking at the Starkware ecosystem, it’s still very nascent but you have some core permanents in place. You have DeversiFi, for spot liquidity you have dydx, for perps, Immutable, NFTs, I guess, what use case or application do you guys think, this is your pitch to builders out there listening to this podcast? What application do you think start where needs, that would be a game changer with the ecosystem that’s missing currently?

Uri (45:25):

So on StarkNet itself, I think there is a bunch of stuff that folks are actually talking to us about. In the context of DeFi, there has a bunch of things that people are trying to do on the gaming front, which I think is just absurdly exciting because unlike… well maybe I don’t know if I’m like, but anyway, it’s fundamentally different from the gaming world as we know it. And so I think that’s very exciting. On the StarkEx front in a few weeks time, we’re bringing DeFi pooling to market. And I think that’s a very, very, very interesting use case. 

And I expect other applications, other than seller or bringing this to market seller network. We expect other applications to adopt this. And DeFi Pulling is a very powerful concept that I think addresses the huge limitation put in place by Layer-One transactions in terms of allowing the long tail of retail investors to tap into this opportunity, which is otherwise permissionless and very exciting and open and borderless and all that.

And then it has these crazy Layer-One transactions to basically limiting this market to the ‘whales’. So what is DeFi Pulling? DeFi Pulling allows users to exist on Layer 2. DeFi remains on Layer-One, where it was born and where a lot of exciting stuff is still being built. And DeFi Pulling essentially allows for the pulling of demand from users for various DeFi services on Layer 2. And from there you’d launch, think of it like a shuttle bus or a commercial flight, as opposed to a Gulf stream.

A shuttle bus that goes from Layer 2 to Layer-One, whatever, to avail or compound or whatnot, transact a single Layer-One transaction representing a very large set of users. And then coming back with those proceeds to layer 2 to be distributed, trustless need to user. Now this is very exciting, because it basically shields users from Layer-One transactions.

Medio (47:42):

And would that have privacy benefits as well for those individuals? Or is that not really part of the feature set?

Uri (47:49):

That’s an interesting point. That in this regard people would have greater privacy than they would have on Layer-One. So it’s not privacy from say the operator of the DeFi-point service, but unlike your transactions to on Layer-One, these would not be documented there. There wouldn’t be a single whatever, 10,000 or $100,000 transaction going to Ave representing who knows 500 users and they’ve pulled their demand on there too.

Tom (48:23):

Okay. Switching gears a little bit, just curious. Do you guys have any interest in attracting the traditional gaming franchises, the colleges of the world, something like Starkware? Or are you more interested in like the new creative kind of crypto native games being built?

Eli (48:37):

Yes, very much so. Very, very interesting.

Tom (48:41):

Is there any blocker to attracting traditional, I guess it’s more educational than it is tech or an onboarding at that point?

Eli (48:48):

Conservatism on their behalf. And also I think we spoke with quite a few of these traditional electronic gaming companies and among those who have expressed interest and appreciation for blockchain, there’s also this issue regulatory and compliance that makes some of them uneasy and worried and slow to move. But from our point of view, we have these skills and capacity to service them. So we very much like to do it.

Uri (49:23):

I want to point out that in addition to that, I think there’s, these franchises are extremely profitable franchises, massive cash cows. Now bringing this to the blockchain space means going from a closed economy, a game economy to an open economy. Now how that affects the distribution of wealth and the game, user demand, et cetera. I don’t think that’s 100% clear to these game studios. So you see all sorts of stuff like doing these on a private blockchain, which is essentially a closed ecosystem. 

That completely defeats the purpose. So at the end of the day, I think that what’s going to happen. And that often happens in various markets with various technologies. Is that the innovation, the massive innovation, like the breakthrough stuff is going to come from new teams, building new stuff. And the big boys are going to follow the lead once they see what succeeds and what doesn’t.

Medio (50:22):

That point was essentially the conclusion we reached back in the spring, we did a post focused on different NFT scaling solutions comparing, what Immutable  was offering on top of Starkware versus polygon versus dapper and a few others. And that was the takeaway for brands that, this was one of the few places where it could meet the scaling needs that they require. If you’re a mainstream brand and you have billions of people around the world that want to interact with this, the private blockchain solutions just simply don’t have the throughput potentially, you need to even be an order of magnitude higher than that. 

And in that regard Starkware is essentially in a league of its own, and that’s hard for them to add later on. Like it’s core, it’s fundamental to the network design itself.

Uri (51:11):

You can’t just slap that on at the the last moment. Yeah. It just goes to the heart of the whole thing. Yeah.

Medio (51:18):


Tom (51:19):

Guys, it’s not often we get both of you guys on a podcast, so I’d like to reminisce for a little bit. You guys are at a critical point where your life, people love you guys for building, just to think back over your journey, I guess, what were the biggest learnings or hurdles each of you had in this process? Like what advice could you give a founder building something as complex as you guys and as impactful? Eli, we’ll go to you first. I feel like you’re ready to go.

Eli (51:44):

No, I was just thinking, I don’t know, like put sunscreen or one of those that I’m blanking.

Uri (51:50):

Sunscreen is important. Certainly [inaudible 00:51:55].

Eli (51:55):

Okay. I’ll just say, but again okay. If you’re in… I was just asked by an academic, like two days ago about like there was this nice result recently research and like, how do you start playing with commercialization? And for that, my tip just was that, the way, at least for being back as an academic, okay, take your theoretical work and maybe find someone or try to program it and just to make it also from the paper onto this thing, but that’s not that relevant for probably for your listener base.

I would say pick a good co-founders and teammates. So, I obviously feel I made a mistake here, but with this guy, but learn from that.

Uri (52:48):

It was back in the-

Eli (52:49):

No, just joking. I don’t know. Yeah. There’s so much on the internet on how to do things well, try to enjoy.

Uri (53:01):

One particular challenge that we’re seeing has to do with there’s the tokenomics sort of create this fog of war, and crypto Twitter is also a tremendous source of noise. It’s also a tremendous source of value, but it’s a tremendous source of noise. And being able to see through all that clutter and say, this is real, this is built, this is crypto Twitter, that was deployed on crypto Twitter versus that was deployed on maintenance.

Being able to see through that and being able to say, well, all that adoption is driven by the shitcoin around that, whatever it is, that’s important. And it’s not easy at times because it takes time for the dust to settle. So I think being slightly older and knowing one another, going back to our formative years, this relates to a bunch of us at Starkware, it helps to manage all that signal.

Eli (54:08):

I want to add another tip that I think is especially important to founders in the crypto world. So we meet a lot of teams where the story is something like this. They had a really good idea and they were three young buddies from whatever college or somewhere, and then they started working on it and they’re really happy being these three buddies that know each other. And then suddenly it just the succeeds. So it’s ballistic and this scaling is building like a robust team was planning ahead and people with experience and managerial experience and so on. 

I think you want to do that early on and factor it in because otherwise, we meet all of these things where you see that they’re like in thrashing mode everything’s going tremendously well, but they don’t have the mindset and capacity to expand beyond the three to four buddies. And so maybe that’s one tip.

Tom (55:15):

That’s great, great advice.

Medio (55:18):

That’s sage wisdom and that’s something, we went through those same growing pains as a company as well, starting a small group of friends and then expanding over time. So far, this conversation is focused on scaling with tech, but I think ending on scaling the human side of things is a good note to end on.

Tom (55:38):

That’s a great point, Medio. Uri, Eli, it’s been incredible to have you guys on, we’re obviously huge fans of Starkware and everything you guys are building. And just shout out for all the hard work you guys have done. Is there any way that or what’s the best way for people to get involved with Starkware? And I guess where you look for maybe developers projects, where they go, who should they talk to?

Uri (55:58):

Sure. So I would sort of split it across. If you’re mostly a reader Starkware LTD on Twitter to read what we’re up to. If you’re a writer namely a coder, and you want to build and go to our StarkNet discord. The link can be found on the Startware Twitter handle, on my Twitter handle. So those I think are the two focal points for information and interest and meeting other like-minded people building teams and all that.

Tom (56:34):

I love that Uri, Eli, thanks so much. Medio our head of research. Thank you so much for coming on to guys.

Uri (56:39):

Thank you for having us.

Eli (56:40):

Thanks, Tom. Thanks Medio.

Show Notes: 

(00:00:00) – Introduction.

(00:00:44) – Guests’ backgrounds.

(00:01:39) – Overview of StarkWare.

(00:05:03) – Coding with Cairo on StarkWare. 

(00:08:14) – Deploying game logic on top of Starknet.

(00:09:48) – The low cost of transactions. 

(00:14:34) – Implementation speed for integrating dApps. 

(00:17:44) – How StarkWare compares to optimistic rollups. 

(00:22:42) – Platforms building on top of StarkWare.

(00:25:39) – What success for StarkWare looks like. 

(00:26:52) – Data availability, sharding, and Validium.

(00:31:08) – Transaction costs on Validium vs. ZK. 

(00:34:19) – Addressing critiques of StarkWare. 

(00:41:30) – Being ahead of schedule.

(00:43:28) – Building a community without a token. 

(00:44:34) – Game-changing applications on Starknet. 

(00:48:07) – Attracting traditional gaming franchises. 

(00:51:06) – Best advice from Eli and Uri.

(00:55:44) – How to get involved with StarkWare.