Antonio Juliano was a fresh-faced computer-science grad from Princeton University in 2015 when he joined the U.S. crypto exchange Coinbase as a software engineer. He only stayed for a year, but he caught the blockchain bug.
Now Juliano oversees one of the biggest and most closely-tracked decentralized exchanges, dYdX, which he founded in 2017.
This article is featured in the latest issue of The Protocol, our weekly newsletter exploring the tech behind crypto, one block at a time. Sign up here to get it in your inbox every Wednesday.
He’s been in the news lately, because dYdX is on an odyssey – moving from a layer-2 network in the Ethereum ecosystem onto its own blockchain built using Cosmos technology. A test network was announced in July, and after some delays, the main network (mainnet) launch is now expected next month.
We spoke with Juliano last week about the transition, as well as some key blockchain issues that the industry is grappling with. Some highlights:
Q: You’ve got the main net launch coming in October, but you had previously targeted September. What’s the status now? And what are the things that you have to do to get there?
Juliano: It’s a pretty big project – building a whole layer-1 blockchain. We’re also building some really interesting novel things into our blockchain. At a high level, we’re building a decentralized but off-chain order book and matching engine, which is going to be way more scalable than anything else that’s existed in DeFi. We’ve been working on this for a little over a year now. And I actually think that’s a really fast timeline to be able to bring this to market. And at the beginning of the year, we’d set the end of September date, and it got pushed back. It’ll probably be sometime early- to mid-October at this point. But I still feel really good about that. Obviously, we want to have a lot of confidence in the security and correctness and stuff of what we’re launching. TLDR it’s pretty much completely built. And we’re just in the final stages of testing.
Q: You have to get your own community of validators, as opposed to other setups where you might just kind of inherit somebody else’s security domain. What’s involved in that? How do you find those validators? Or what’s the challenge there?
Juliano: So it actually hasn’t been that challenging, surprisingly. And it’s not really something we’re even involved in. We’re really just focusing on building the open-source software. And I think one of the things that enables that for us is, dYdX is not a new project, right? We’ve been around for six years. We have around a billion dollars a day that’s traded on dYdX. The brand is pretty strong. And I think for the people that are validators, and this has actually become a pretty robust community within Cosmos and some other appchains, where there’s lots of people that have made a business out of running validators. To have a project like dYdX, come in with a really big user base with a really big brand already, it’s very likely, without very much, or any effort from our part, or the rest of the dYdX community parts, that people are going to want to validate, because there’s incentives to do so. So that’s kind of how we think about it. We’re probably going to start with something on the order of 50 to 100 validators and then scale up from there.”
Q: Can you talk about how the new dYdX chain might scale your project versus your prior versions of the protocol?
Juliano: We do need a really high amount of scalability, which is kind of the main reason that we’re building what we’re building. And the reason for that is, obviously, we can’t build something like this on Ethereum. Maybe that’s not even close to scalable enough. We considered building on some other chains like Solana or potentially other L2’s. We decided against Solana because it doesn’t actually have enough scalability for us. Not to claim that our scalability is that much higher than its allotted high level. I think it will be higher because of the off-chain order book. But the other problem with that is that we don’t have all of Solana to ourselves, right? So even if Solana it is every bit as scalable as the dYdX chain, for what we’re building, to share it with other people, pay transaction fees, and that’s not really the user experience that we’re looking for. And then on the layer-2 side, there’s kind of two problems with that. First of all, the layer 2s are just not even close to scalable enough yet. And we actually know that very well. Because today we are pretty much the biggest app that’s been built on layer 2 to date. We’ve been running on StarkWare, not Starknet, but a proprietary deployment of the StarkWare protocol that they built for us. And that’s been going really well, and that’s been an awesome partnership. But the main problem is that it’s not scalable enough. And because of that, we have to run components of our architecture off-chain.
Q: Can you break it down in terms of some numbers?
Juliano: So, L2’s right now can kind of process on the order of 100 transactions per second or so, which is great. That’s kind of a 10x improvement over Ethereum itself. But we need more on the order of 1,000-plus transactions per second, just because of the nature of the product that we’re building. So obviously, decentralized exchanges were targeted at financial derivatives. And our main customers are institutions and pretty advanced crypto traders, who need a really high amount of scalability and need an order book to be able to trade on. That’s one of the key things to understand about dYdX – that while most decentralized exchanges these days are automated market makers, dYdX is not, at least, so far. We’re an order-book-based decentralized exchange. And that requires a really high amount of scalability.
Q: Why is that?
Juliano: Just think about what’s going on on the platform. We have all of these trading bots, market makers. Individuals are running that are placing and canceling orders all the time on this order-book-based architecture. That’s just kind of how exchanges work before. When you’re placing and canceling these orders, they’re normally not getting filled. Maybe 1% of the orders on the chain will get filled. And market makers really want to not have to pay transaction fees, if they’re just placing an order and it’s not getting filled. Why would you want to do that? That’s a terrible experience. So that was kind of the key insight that we had that brought us to the conclusion of, why we’re building what we’re building is that, first of all, we need a really high amount of scalability for placing and canceling orders. But we actually don’t need that high of a scalability yet, at least for the trades themselves. Maybe 1% of the trades match, and those actually go into consensus.
Q: How does it work?
Juliano: What we’re building on V.4 and the dYdX chain is what I term, a decentralized, but off-chain order book. And it is decentralized, because it’s run by the validators themselves, kind of think of it like, you know, the mempool in blockchains. In general, we’re effectively using the mempool to run this order book, but we’re just customizing the mempool in a way that makes it really efficient for running dYdX. So people can place and cancel orders all the time, they’re going to this decentralized network of the validators themselves. But the cool thing is, they’re not actually going on-chain until they’re matched. And that’s awesome, right? Because again, like only 1% of the orders actually match. So that gives you pretty much 100x scalability right there. Because these things on the order book, you don’t have to come to consensus on them. And that’s just a really hard computer science problem, right? Like, how do you scale blockchains, we have some thoughts on that. But we’re not going to be 10x better at scaling blockchains than everybody else that’s out there right now. But we can kind of have these key insights and build something really custom for what dYdX’s unique needs are.
Q: You were saying you need on the order of 1,000 transactions per second. What do you expect to actually achieve, post-launch?
Juliano: When most general purpose chains do benchmarking, they’ll look at, oh, maybe I could send you ether or send you a more simple transaction than something that’s more complex, and it might take more gas. For us, trading is trading and we know what it’s going to be. And we’ve tested with the network of, I think, on the order of 50 validators or so, and achieved 2,000 transactions per second so far, and that’s pretty cool. Because, remember, again, there’s no gas fees involved in this. So you can kind of get 2,000 transactions per second with no gas fees. And we’ll continue to improve from there if there’s a lot of low hanging fruits in the Cosmos SDK and Tendermint, some of which we’re working on ourselves to improve, some of which the Cosmos community is working on to improve, but that’s actually the level of throughput that dYdX V.3, which is the current version of dYdX, has right now. It’s on the order of 1,000 per second.
Q: So why did you need more?
Juliano: We want to grow, right? So I think if you were to look at a Coinbase, or a Binance, I don’t have the stats in front of me. So we’re topping out StarkWare, we’re at sort of around 1,000. But let’s go one level deeper and kind of understand that a little bit more. So the 1,000 transactions per second that I’m quoting right now is just on our backend. That’s before you even hit StarkWare, and it’s just the order book that’s running on our servers in the cloud on AWS, then again, kind of remember the main metric, which is 1% of the orders actually fill. So if we have 1,000 orders placed per second, well, that means only 10 orders per second or so are actually filling. And that’s the amount of throughput that’s actually hitting StarkWare right now is probably on the order of 10 to 20 per second.
Q: With dYdX you’ve obviously got this more kind of hybrid, centralized/decentralized exchange model.
Juliano: Yes, definitely, I would characterize dYdX V3 as a great example of a hybrid exchange. There are some centralized components, and decentralized components, the whole thing is noncustodial. That’s maybe there are tradeoffs that you’re making from a decentralization perspective. And I think that’s worked pretty well, for us so far. And I’m really glad that we did take that path. I think it allowed us to scale a lot, allowed us to iterate more quickly, find product market fit and build a product that our users really want. But the plan behind dYdX and I would argue most if not all DeFi platforms, I think, or at least the ones that that seem reasonable to me, is to become fully decentralized at some point or another. So that’s how I’ve always thought about it as it’s just kind of a step along the path. We do really want to build something that is fully decentralized, just because you can’t get all the benefits of DeFi unless you are fully decentralized. And that’s kind of the main thing, if you remember one thing about the dYdX chain, or V4, it is fully decentralized, whereas V3 was not.
Q: In the past year, community sentiment towards those sorts of hybrid models has shifted. Let’s hear your take on how you pitch dYdX within that framing, and whether you’ve seen sentiment changing.
Juliano: There’s certainly trade offs to both and there are honestly some people that care about decentralization, and some people that don’t. I’d say institutions in some ways care less than retail about the centralization, not zero. One of the really cool things that’s happened on dYdX and some other platforms recently, is institutions like Wintermute, and some other progressive trading firms have taken a really big stance in governance. And I think they actually value that a lot, kind of being able to be on a platform that’s totally transparent, they know is fair, and they know they have some control over going forwards. Like that’s just really, really different than trading on a Binance. Even if the Binance is every bit or better of a product experience for them. Just kind of like having this control. And this transparency is really valuable as well.
Q: What are you looking for in terms of the composition of your validators, and whether that’s more organic like Ethereum and Bitcoin or more institution-based?
Juliano: At least the state of the world right now on Cosmos is that the validators are indeed known entities and pretty much entirely, they’re more institutional, smaller institutions like institutions and kind of the crypto sense. A lot of these shops are, OK, maybe it’s a company, but it’s just a few people running one of these things across a couple of different networks. And that’s where we’re at right now, I think there’s a lot of development effort in Cosmos that’s going into increasing the number of maximum validators that you have. And the reason for that is, so there’s just kind of this relationship with how Tendermint works, that the more validators you have, generally, the less scalable the chain is. And as much as we can improve those things, then it’s like, we can have more validators with the same or better level of throughput. And that’s something that will take time.
Q: Antonio, you wrote this spicy tweet on Aug. 24 about single sequencers in the Ethererum layer-2 landscape. What was that all about?
Juliano: I do strongly believe that layer 2s will get to a point where the sequencer is a network, rather than just a single operator. And that will make the sequencers much more decentralized, censorship-resistant. But the problem with having a single sequencer on layer 2’s right now is, the single sequencers are not censorship-resistant, right? It’s just like, literally a server, and the server can refuse to process your transactions if it wants to. Like, I’m not trying to pick on anyone, like literally all of them are like this, but let’s just take like Coinbase or Base as an example. Coinbase is literally running a server that is the sequencer for the Base chain. And if they want to reject your transaction, for whatever reason, they can do that. And, you know, I think they’re trying to be as impartial as possible. And that’s awesome. But that is not totally decentralized, I may say. And just to prove the point that I kind of understand most of the arguments at least, OK, there’s a way to get around the censorship-resistance switches. Effectively, you can go back to layer 1, and send a transaction to layer 1 and be like, Oh, Coinbase, you have to include my transaction now because I went to the layer 1. But that’s a terrible product experience, right? Like maybe that works for some edge cases, or if like the less than 1% of the time when, for some random reason somebody’s being centralized by a sequencer. But if there’s a good amount of opportunity for a sizable amount of the transactions to be censored on a given sequencer, for whatever reason, then this may not be something that you want for your network as much.
Q: I would be interested in your thoughts on these recent CFTC settlements announced with some other decentralized exchanges?
Juliano: Yeah, I mean, it’s something that we obviously pay a lot of attention to. So I think we have had, like, we have had conversations with the CFTC for six years now, which is the entire time that dYdX has been around. And we haven’t had anything happen with dYdX so far, which I think has been a really great thing and I think speaks to the amount of compliance and thought that we put into being compliant with regulations. It’s something we think a lot about: How can we be innovative but still within the bounds of the law, both as they are today, and then how can we be potentially change the law through policy going forward? So I think that’s how you try to play it right now. In some ways, the laws are very clear. But in some ways, a lot of this precedent is still being established within crypto. And that’s a great thing, right? But, you know, you can look at maybe a different example of Coinbase and the SEC. And Coinbase is now openly taking the position that they disagree with the regulators.
Q: What are some of the key differences for why those sorts of allegations might not apply to dYdX?
Juliano: I’m not a lawyer. And I can’t talk about totally transparently everything we’re doing internally, right? There’s a lot of stuff that we do internally, from a compliance perspective, that does make it fundamentally different. I mean, the most important thing to note is that dYdX has never offered derivative products to U.S. customers. That is pretty much like the main thing right there, right? And we have had these conversations with the CFTC. But so far, obviously, we haven’t gotten to a place where we feel comfortable offering our products to U.S. customers, at least with the laws as they are right now. And that’s why we’re trying to change them in the future. So users in the U.S. can get access to these safer products and aren’t exposed to trading on these unregulated exchanges offshore, because none of the regulated ones can serve their needs. So I think that’s kind of the fundamental difference is that we’ve been really thoughtful and careful about this. We were fortunate to be well funded, fortunate to have worked with all of the top blue chip investors in the space. That doesn’t solve all your problems, but at least gives you access to the best resources, right? Like we have, like the best lawyers. And I think it is a little silly, but you know, this is kind of the way it is that, a lot of the details matter in these cases.
(This interview has been edited for brevity.)
