"June 2024 Interview with Jeff: We Don't Compare Ourselves to Any Project, We Innovate Continuously from First Principles"
Question: How has Hyperliquid, as an order book-based decentralized exchange (DEX) and L1 layer project, managed to stand out in the highly competitive perpetual contract DEX space? How has your team's background helped in this?
Jeff: On the surface, building an exchange or DEX seems like a simple problem; everyone knows how it should operate, and an order book isn't rocket science.
But details determine success or failure. I believe that at least a team with a strong background in quantitative trading is necessary to build a good exchange, even a centralized one.
When it comes to building a decentralized exchange, a deep understanding of the underlying infrastructure and how to scale solutions across the entire chain is essential. There are at least hundreds of items that need to be checked off when building an exchange. If any one of them is not done well, such as a poor user interface, bad API, or insufficient liquidity, traders will not use your product.
Our team comes from a quantitative trading background, so we can tackle many of the more challenging aspects, and we also strive to identify shortcomings and improve them. This requires a lot of hard work, listening to user feedback—there are no shortcuts or secret formulas; it’s about continuous iteration and product development, never settling for any flaws.
Question: What conditions must a decentralized perpetual contract exchange (DEX) meet to reach the same level as centralized exchanges or increase market share?
Jeff: I think this is a puzzle for everyone. In my view, whether centralized or decentralized exchanges, innovation is necessary. Just like Binance, if they look back at Bitmax and only think about how to match what they do, users won’t switch over.
So, we have innovated in almost every area possible on Hyperliquid. In many aspects, it is actually better than centralized exchanges, and in other areas, it can reach the same level. Factors like liquidity, throughput, and responsiveness to market makers are very important. There are no shortcuts; we can only build good technology, and that’s what we focus on day in and day out.
Question: You have the hlp Vault and personal Vaults that can be created by individuals. I think this Vault system is quite uncommon; it seems like a good marketing tool. Did you launch it because you believe this approach will be successful? Why did you decide to introduce it?
Jeff: This is actually quite an interesting matter. We have built a Vault primitive at the L1 layer; it is essentially a foundational component. People think of it as a product, but it is more like a signature primitive. The benefit of building L1 on Hyperliquid is that we can customize every aspect of it.
The cool thing about the Vault primitive is that you can create an address and sign on its behalf, but you don’t have the authority to move the funds in that address. This address ultimately becomes a foundational component for copy trading. Initially, we built it just to issue hlp, but then we thought, since it’s already built, why not let everyone use this primitive without permission? Maybe they can come up with great strategies.
At first, I thought that Uniswap pools (especially V2 pools) could be copied as a simple Vault strategy, and it might perform better because it wouldn’t be affected by arbitrage like Uniswap.
But in the end, no one cared about that point; instead, everyone utilized it in their own ways, leading to many cool and innovative ideas. Honestly, this exceeded my expectations. It took several months for people to realize what they could do with it. So sometimes, when you build the right foundational components that can’t be done elsewhere, people will naturally use and promote them.
Question: Your cross-chain bridge has one of the most seamless user experiences I’ve seen. How did you achieve this? What choices did you make regarding trust assumptions? Are there plans to add cross-chain bridges from other chains? What will the situation look like in the future?
Jeff: We have thought a lot about the cross-chain bridge, and we even feel we might have spent too much time on it. The final product is visible to everyone, and it is open-source, having undergone two audits. The code is about 2000 lines of Solidity; it is a custom cross-chain bridge.
Regarding trust assumptions, it is secured by an arbitration group of L1 validators, and there is a system in place. If a vulnerability is exploited, a broader group of monitors (similar to the dispute resolution mechanism behind the Arbitrum bridge) can lock the cross-chain bridge, and it requires signatures from a cold wallet of the same group of staked-weighted validators to reach the arbitration threshold to unlock it.
Seamlessly integrating it with L1 to achieve a good user experience took at least a quarter or even half a year of iteration and refinement. We also considered many other solutions, such as IBC or other customized solutions, but realized that the cross-chain bridge is the first thing users will encounter. If we outsourced it to other projects or protocols, we might not be able to optimize it for our needs. Our experience with other cross-chain bridges is that they often fall short in user experience, so we took on the task of building the entire cross-chain bridge ourselves.
As for the other question you mentioned, with the launch of Hyperliquid's EVM, an important reason is to support more diverse and institutionally-oriented cross-chain bridges.
One thing we are particularly excited about is being able to support native USDC. Currently, this custom cross-chain bridge has a bit of a bottleneck in attracting institutional funds into Hyperliquid. For example, some institutions may not want to use Arbitrum, which is currently a single point of entry into Hyperliquid L1.
But once the EVM is online, CCTP will allow anyone to burn USDC on the source chain and generate it on Hyperliquid. We also want to address this issue for other spot assets in a more universal way, which is a significant motivation for us to invest effort in developing Hyperliquid EVM.
Question: It seems that a few months ago, Hyperliquid shifted towards expanding L1 functionality, no longer just limited to being a perpetual contract DEX. Was this part of the initial plan? What is the reasoning behind it?
Jeff: This was definitely not part of our initial plan. Initially, we just wanted to do something helpful for the entire crypto community. In 2022, we traded on many centralized exchanges and DeFi platforms, gaining a clear understanding of everything happening at that time, and we used many newly launched protocols, only to realize that they were mostly not very good.
After the FTX incident, we strongly felt that the space lacked true builders. FTX was supposed to be built for traders but betrayed everyone’s trust. At that time, people thought it was DeFi's shining moment, but we had been trading in DeFi and felt that it couldn’t compare to FTX at all. So we thought, what are we best at? That is building a truly good decentralized exchange that operates entirely on-chain and can genuinely scale to the size of Binance, but we also needed to be grounded and start from the markets we were familiar with to attract users.
That was the initial plan—thinking about what we could do and contribute. But gradually, as we worked and received positive feedback from users, our ambitions began to grow. Initially, we thought we just needed to do what we were good at, but later we realized we might need to delve deeper into building more things. At a certain point, we felt that the other work we were doing could also be released to benefit others.
Moreover, many of our users are actually great developers who want to create various dashboards and launch various products on Hyperliquid. Initially, we thought we would mainly focus on being a perpetual contract DEX, but then it suddenly dawned on us: since there are so many talented builders and community members, why not provide them with the tools to do everything they want to do?
Question: Many people think that most order book protocols are fully on-chain, but in fact, many move part or all of the order book and matching engine off-chain. How do you manage to keep everything on-chain? What is your approach, and won’t you encounter bottlenecks?
Jeff: Initially, we used Tendermint because it is a well-tested consensus mechanism and easy to modify. After some adjustments, it could handle about 20,000 transactions per second. At that time, there weren’t that many orders, so it was manageable. But now, Hyper L1 has completely migrated to what we call Hyper BFT (which isn’t a consensus mechanism we invented; many L1s have been using related consensus research results over the past five or six years. We just read those papers, spent half a year reproducing them, and turning academic results into practical applications isn’t easy, but once we had the core idea, we manually implemented it).
We pay close attention to all underlying details to ensure that throughput does not decline during actual implementation. Looking at it now, consensus isn’t even a bottleneck; I believe that if execution speed increases, consensus can keep up. Currently, the bottleneck is primarily execution, followed by hardware and network bandwidth. However, I think in the future, the bottleneck may return to consensus, but if we reach that scale, something like Binance could run entirely on-chain.
Question: Can you explain the differences between Hyper BFT and Tendermint?
Jeff: The core difference is that Tendermint is based on earlier research results, and at that time, some new excellent research had not yet emerged. I think those new results may have also been inspired by Tendermint and other algorithms; after all, we stand on the shoulders of giants. So, Hyper BFT can be seen as an iterative improvement.
A key difference is that Tendermint has some built-in waiting mechanisms related to how its algorithm operates, involving time scales. In contrast, the Hyper BFT algorithm itself is more asynchronous. In the multi-step process, there is no need to set how long to wait before proceeding to the next step; as long as enough votes are obtained (in this case, more than two-thirds of the stake), one can vote on this block and move to the next step. In this network, any validator does not need to wait for a specified duration in the normal flow of the algorithm. Therefore, when the network runs smoothly without timeouts, the algorithm can advance at the fastest speed of information transmission between validators. This is a significant core difference and essentially means that block time has no lower limit.
Another major difference is that Tendermint was implemented earlier as a result of extensive experimentation. Initially, it was more of an academic implementation, focusing more on research, so there are many small details in the code that may seem meaningless when viewed individually, but collectively they can create a significant difference in production performance.
Question: You mentioned that many people are developing on Hyperliquid. Do you allow permissionless smart contract deployment? Are you looking forward to people launching new products? For example, early GMX gained a lot of attention because people built new products on it. What are your thoughts on this?
Jeff: We are very open to letting the community do things in the way they see fit. I think there is a good side to general-purpose platforms, even though I have often said they are not ideal because it may not be possible to build specialized things with them in specific scenarios. However, the permissionless smart contract system does have a cool aspect: it doesn’t require us to dictate what people should build. We provide them with tools, and they will build on their own, which I think is powerful.
Our vision is to combine the two, which is quite unique on Hyperliquid. Perhaps we should explain this more to everyone because people think we launched the EVM just because it's cool and allows for building things. While some may want to build on it, the real reason is that it enables Hyperliquid L1 to evolve from a chain focused on specific applications into a more comprehensive application platform, primarily for spot and perpetual contract trading.
To facilitate perpetual contract trading, we want to make on-chain operations easier, and the EVM can solve this problem because many cross-chain protocols prefer to integrate with EVMs, which is the first thing they need to ensure.
In terms of spot trading, issues like bridging standard spot assets such as ETH, BTC, and SOL from other chains can also be resolved through cross-chain bridges. More importantly, you definitely want native assets on-chain. Solana has done well because many new builders choose it and build on it. When your native application is on Solana, the most natural place to trade native assets is on that chain rather than bridging to another place. Hyperliquid is the best choice for this, so there is a natural synergy in building.
So, in a sense, the EVM is a generalized solution for this already expanded system. We are not starting from something general and then expanding; rather, we are starting from a solution that already has scale. We are confident that it can scale on-chain to the size of Binance or even larger, and then incorporate generalized functions.
Specifically, we do not expect anyone to replicate what Hyperliquid L1 currently has natively on the EVM. After all, Hyperliquid users may prefer it because they like the order book, while something like GMX or Solana may not be as interesting.
However, there are many long-tail financial needs that can be built in a permissionless manner on the EVM. I think lending protocols would be very cool; there has already been a lot of work done in this area on the EVM, with many large projects and high locked value. Users often ask if they can use ETH as collateral to trade, either for arbitraging ETH itself or trading other coins that use USDC as margin. In this case, there is no need to build these functions into L1 (which is a big project); a composable lending protocol can be directly implemented in the EVM. I think that would be very cool, and there are already people expressing interest in building such things, which I look forward to.
Question: If you allow other foundational components to be built on L1, would you be concerned about congestion or block space issues?
Jeff: I think there is a significant misunderstanding. We are not looking to build a high-performance EVM or anything like that. I believe the demand for EVM on Hyperliquid will be relatively small, and I hope it stays that way. If you look at Ethereum and see the transactions that end users are making (like Uniswap trades, USDC transfers, and L2 blocks on the L2 side, which are simple financial operations performed by end users), if you add these up, I don’t know the exact proportion of Ethereum block space they occupy, but I would guess it’s quite high.
So when all of these are moved to a native application, I believe the real demand for EVM space will decrease. This is actually an alternative way to scale the EVM that hasn’t been discussed much—putting things that shouldn’t be built on the EVM or any virtual machine, which should actually be natively built into the protocol, into a native environment. This can alleviate congestion, and that’s our vision.
Of course, if there is indeed a huge demand for this EVM, it doesn’t mean we can’t scale; we just need to see evidence that such a situation will arise before we make corresponding plans.
Question: This Monday, you announced that Hyperliquid L1 will support native EVM. One exciting feature of this integration is that hip1's off-chain assets will be able to undergo atomic transfers with the corresponding ERC20 contracts. What can users expect from this upgrade? How will it change the way users interact with the product?
Jeff: What you mentioned is one of the examples we provided, and I think it’s the most obvious example showing that this EVM is not an isolated new feature. Soon, people will be able to deploy hip1 and hip2 on the mainnet without permission, and we are very excited about this. We believe it will be very cool for many developers.
With the EVM, developers will have more confidence that they can do anything they want in the future.
Now, they can launch hip1 assets with corresponding Hyperliquid liquidity (which is a strategy for providing on-chain order book liquidity). Once the EVM is online, the deployers of hip1 and hip2 will be able to specify a corresponding ERC20 contract in the EVM state. The EVM state and other L1 states are composable; they reach consensus to form a unified state. This will create a mechanism that allows the token balance of hip1 and the token balance of ERC20 to be interchangeable in a certain sense. This is not like a cross-chain bridge; it’s an atomic transfer.
So for developers, if they are considering using the EVM (and I think many developers will), this is a strong attraction for them because two important points for developers are which chain to build on and how to decentralize the governance tokens or gas tokens and other tokens they need to issue:
The first question is whether to choose EVM or something else.
The second question usually involves going to places like Binance or Coinbase and begging them to list their tokens. Exchanges have a lot of power in this regard, and I don’t think that’s a good mechanism.
Hyperliquid is a completely open on-chain trading platform, and these can be automatically realized on the same chain you are building on. I think that’s very powerful. So if we can atomically combine Binance, Ethereum, or other scalable solutions into a whole, that is the vision of Hyperliquid EVM.
Question: Currently, traders using L1 do not have to pay gas fees. How will decentralization and the introduction of native tokens affect costs? What challenges do you think you will encounter in the process of introducing and attempting to decentralize the validator set?
Jeff: L1 does indeed need a native token, but it doesn’t necessarily have to be for paying gas. I think people often focus too much on gas. In fact, L1 needs a native token mainly for staking. Without a native token, it’s impossible to run a proof-of-stake mechanism. In my view, this is the only real function that L1 tokens need to have, but it’s a very important point that cannot be overlooked.
Currently, there are no plans on Hyperliquid to require trading with L1 tokens. Of course, that situation may change, and perhaps new mechanisms will emerge that make it more reasonable to associate trading with L1 tokens, but for now, there’s no reason to believe that will definitely happen.
The issue of trading without gas fees is not a temporary measure; it’s not that we should have charged gas but didn’t. Rather, L1 itself is highly customized, and we can achieve the same effect as charging per transaction with a native gas token through other mechanisms. For example, we can ensure payment for each unit of throughput or each unit of state expansion through a combination of fill rates and trading fees, which is essentially the role of gas.
So, at first glance, Hyperliquid’s system may seem like it’s covering up the problem, but it’s actually well thought out. There may still be some issues, but I believe the basic idea is reasonable, and there is a lot of explanation about the principles in the documentation.
This will indeed require users like market makers to adjust their strategies slightly, but by now, everyone basically knows how to adapt, and I think this is a fairly scalable solution. Another good feature is that for first-time users of Hyperliquid, they don’t have to buy a token they didn’t intend to hold just to use it.
Question: The Hyperliquid points program ended earlier this month, but Hyperliquid still maintains a top-three position in daily trading volume among perpetual contract DEXs. Unlike other points programs, you didn’t attract attention through multipliers or other collaborations and clear rules. What do you think has attracted the right users? Or aside from the product itself being good, is there any secret?
Jeff: That’s a good question. We actually didn’t think much about how to market the points. We believe in simplicity; points are just a number. They have no intrinsic value; they are used to indicate how much users have contributed to the protocol. We think it’s straightforward and self-evident, and we don’t want everyone to focus too much on this number. We still hope the product itself can speak for itself.
However, it does serve as a talking point that can attract attention. For example, if someone brings up points, others might think, “Oh, I’ll check out this product.” Overall, I think this approach has been effective; people generally like the product, and we are quite satisfied with this outcome.
I think one significant risk we took that paid off was making the points completely opaque. I know many protocols launch points to inflate metrics. I don’t want to comment on whether that’s good or bad, but if you’re going to do that, it’s best to have clear rules; otherwise, you won’t achieve the desired effect. For example, if you tell users they can earn one point for every dollar in fees paid, they will have a formula in mind and can precisely inflate the metrics.
We strongly feel that this is not a good idea. We don’t care about short-term metrics, so I think taking this risk is worthwhile because a significant portion of users, I firmly believe, if you just look at the data, they account for a large share of trading volume in most other perpetual contract DEXs and general DEXs on-chain. With such a large group of users, they are quite fickle and profit-driven. It’s not that they don’t add value to the protocol, but we believe that from Hyperliquid’s long-term vision, they are not that valuable.
I feel that the rules you mentioned, which are clear and can inflate trading volume or activity metrics, are usually aimed at attracting funds or venture capital, etc.
Question: What considerations led you to choose to launch this protocol yourself rather than raising external funds? How did you manage to launch this protocol on your own? After all, your fees are quite reasonable, not like GMX a year ago, where your opening and closing fees were only about twice those of other protocols.
Jeff: Well, I want to clarify the last point first. In fact, our development team does not take any fees from the protocol; all fees are allocated to the insurance fund and primarily to hlp.
That’s right; the fees are low. We firmly believe that in the early stages of DeFi development, where DeFi is still more difficult to use than centralized finance in many ways, low fees are a basic requirement to attract new users. This is a product-level decision.
That said, this protocol has actually collected a significant amount of fees. If it were a commercial project, it would be very profitable, but it’s not a commercial project. I think we need to clarify this point more explicitly. We also occasionally mention this: the team has never taken even a penny in fees because we have a strong long-term belief in this project. We don’t need short-term funding, and we don’t want others to say things like, “This internal team made a lot of money from these fees.” There are quite a few negative comments like that.
I think many protocols have made such choices and have to endure these criticisms, so this is how we operate.
In terms of fundraising, the situation is similar. I think there is a lot of investment interest, but Hyperliquid is very focused on the long term, aiming to become the platform where all financial activities occur. This is a very long-term vision, and there are many obstacles. I believe a crucial factor in determining the future of the ultimate financial platform is that the underlying layer has no internal stakeholders. I'm not saying that venture capital is without value; I think venture capital has indeed contributed significantly to economic development. I don't want to downplay venture capital; it's just that Hyperliquid has a unique opportunity to be more like Bitcoin. I think if Satoshi Nakamoto had done a presale or if venture capital had participated in his initial mining setup, Bitcoin might not be what it is today. I'm not saying we should completely compare Hyperliquid to Bitcoin, but I believe being a neutral platform is a prerequisite, and our ambition is that large, so we need to ensure we achieve this, which I think is very important.
- Question: From a more macro perspective, what are your thoughts on the long-term future of multi-chain, dedicated chains, and whether there will ultimately be one chain that prevails in this field?
For example, in the perpetual contract space, Vertex Edge and Hyperliquid have completely different approaches. Vertex Edge accepts the premise that there will always be a large number of users across different chains, so it tries to aggregate liquidity through off-chain matching, allowing deployments on more chains to bring more liquidity to each chain. In contrast, Hyperliquid may take the opposite approach by developing its own chain through increasing use cases and more protocols, integrating users onto one chain. What are your thoughts on this?
Jeff: I have to say that I don’t have much time to think deeply about such macro issues, so I may not be the best person to answer this question. However, my personal view is that the reality lies somewhere in between.
I believe financial activities should occur on a composable chain, but not everything needs to be like that. For example, if there is a game where items can be traded and millions of users are using it, I really don’t think it needs to compete with Bitcoin derivatives. Bitcoin is a world reserve asset, and their importance is on completely different scales. I think it’s perfectly fine if they don’t combine because there’s no reason for them to do so.
However, I think it’s not good for financial activities to be scattered across many different chains. Cross-chain bridges carry significant risks. But for now, it’s okay because the crypto space is still in its early stages, and I think, generally speaking, teams are acting with good intentions. Even centralized exchanges, overall, want to do good. But this cannot be the ultimate solution. If one thinks centralized exchanges are bad, then a similar skepticism should be held towards cross-chain bridges because they just add another layer and still involve many trust assumptions. They are not automatically composable with other things.
So, as far as Hyperliquid is concerned, at least up to now, we have positioned ourselves as striving to build good products for DeFi. It’s clear that the crypto space has found product-market fit in DeFi, and it’s also clear that even if crypto has no utility in other areas, it is a better system for finance—more accessible and more equitable. In a world where inequality is worsening, enabling financial activities to occur on-chain is becoming increasingly important. Therefore, within this relatively narrow scope, I strongly support a single-chain approach.
- Question: What do you plan to do on Hyperliquid in the next 6 to 12 months? How do you view the competitive landscape of Hyperliquid now and a year from now?
Jeff: Well, we are a very small team, and people are often surprised to learn that we actually have only five engineers, with fewer than ten people in total.
We are not like the common teams in the crypto space that have grand, long-term visions, raise large amounts of funding, and set multi-year roadmaps. I think that approach is pretty cool, but it’s not our strength. We are more focused on the next steps we need to take. I feel confident that we are moving in the right direction, but we won’t lay out a plan with hundreds of steps.
So in the next 6 to 12 months, I think the EVM should be online by then, and the network should be decentralized. We are quite optimistic about these two things, but we also know there is a lot of uncertainty surrounding them. So if any issues arise during this process, resolving those issues may become our top priority.
If no problems arise, I think we will do our utmost to continue promoting the development of Hyper L1 and its surrounding ecosystem, making it a place where financial applications are willing to gather and build around. So regardless of what pain points developers encounter or what issues users, market makers, etc., face as we scale, whatever happens will be our focus.
Ultimately, it’s always about the users, and the definition of users is constantly changing. Initially, it was traders, then traders and market makers, and now it includes traders, market makers, and developers. Whatever they need, and whatever we can help with, that is our primary task.
As for competition, I may not have particularly profound insights. We actually haven’t thought much about competitors, and we’re not very clear on who our competitors are. I don’t think there are many projects doing what we are doing. Typically, that might be a bad sign; if you start a company and no one is competing with you, it might mean your idea isn’t viable. But I think Hyperliquid is currently doing something very general yet focused on core products, which is quite unique.
One metric I often look at is the ratio of perpetual contract trading volume on Binance as the denominator and Hyperliquid’s perpetual contract trading volume as the numerator. I have to say, I’ve always been impressed by Binance and other top centralized exchanges. They do one thing very well: putting users first and retaining them. Take Binance, for example; after going through so much, users still trust and use them. I often look at trading volume data and am always struck by it.
But from our perspective, we are building something different. DeFi is still in its early stages, more immature, and for many people, it’s not yet time to switch over. We are aware of this, so we think we can learn from our competitors, but we will also stick to our own path.
I believe that what Hyperliquid has done well so far has mostly not come from studying competitors and imitating them, but rather from starting from first principles and coming up with something new. Whether it’s the Vault primitives, the permissionless way of listing spot assets, or the gas-free perpetual contract order book (which no one thought could be done before), and the liquidity Vault that makes it easy for anyone to participate in market making—these are all innovations and highlights that Hyperliquid is now known for.
So I think our goal is to continue innovating. I believe this is the true way to add value to the world. If we see competitors doing well and then just do slightly better than them to capture some of their market share, I don’t think that contributes much to the overall advancement of DeFi.
END
Experience the product and get discounts through this link; you can register in seconds with just an email login.
https://app.hyperliquid.xyz/join/DY
Original interview video link
https://www.youtube.com/watch?v=cxyUtPoC1-E&t=110s&ab_channel=0xResearch
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
