Dialogue with the Founder of Starknet: Looking Up at the Stars, Keeping Our Feet on the Ground

Author: Jack, Rhythm BlockBeats

At the recently concluded Montenegro Ethereum EDCON conference, ZK became the hottest topic. From the 19th to the 23rd, over the course of five days, whether in individual speeches or roundtable panels, the discussions among guests mostly revolved around terms like ZK Rollup, zkEVM, and ZKML. In the eyes of the "Ethereum core circle," ZK has already become the ultimate new narrative that can rival the DeFi innovations of 2019 and drive the next wave of industry explosion.

For a long time, the main challenge facing ZK has been its technical implementation, which has led most people to perceive it as a "solution for the future." However, this year, before the Arbitrum ecosystem had fully heated up, ZK Rollups, as technical alternatives, were already knocking on the door. One of the most noticeable feelings this year is that various "ZK projects" have suddenly increased. Developers have started to FOMO into ZK technology, with ZK transfers, ZK identities, and ZK board games; any project related to privacy is sure to have "ZK" in its name. This flourishing scene is undoubtedly attributed to the concentrated efforts of ZK Rollups like Starknet and zkSync, as well as zkEVM projects like Scroll and Polygon. Behind these pillars is a man named Eli Ben-Sasson, a giant from Israel.

Eli (@EliBenSasson) is one of the earliest researchers to combine ZK technology with cryptography. He is the founder of Zcash and StarkWare, as well as the creator of zk-STARK. This technical standard is the crucial technological cornerstone for today's Starknet and the entire crypto ZK track. BlockBeats met this legendary figure during the Montenegro EDCON, and in nearly an hour of conversation, Eli displayed strong technical confidence, revealing his views on STARKs, SNARKs, and the Rollup track, as well as his expectations for the Starknet ecosystem.

On Optimistic Fraud Proofs, zk-STARK, and SNARK

Many people know Eli as the co-founder of StarkWare, but few know that he is also the creator of the key technical standard zk-STARK. Long before Ethereum, Eli began to combine ZK technology to solve blockchain scalability issues and had in-depth discussions on this topic with Vitalik, who was then an author for Bitcoin Magazine. StarkWare is also one of the few projects that Vitalik has participated in investing.

Eli Ben-Sasson speaking at StarkWare Sessions 2023, image source from StarkWare

As a technology-driven entrepreneur, Eli has absolute confidence in ZK technology and the mathematical principles behind it. So how does he view the "ZK rival" Optimistic Rollup? What are his thoughts on the future of zk-STARK and zk-SNARK?

BlockBeats: Hello Eli, first please briefly introduce to the readers why you entered the cryptocurrency field and why you founded Zcash and StarkWare?

Eli: I was once a professor of computer science, doing theoretical computer science research. People in mainland China may have heard of Andrew Yao from Tsinghua University, who is an important promoter of theoretical computer science, a Turing Award winner, and the inventor of secure multi-party computation. I entered this research field many years after him.

In 2008, I began researching zero-knowledge proofs (ZKP). In May 2013, ten years ago, I attended a Bitcoin conference in San Jose, where I realized that the blockchain field was the best application scenario for my theoretical research and mathematics. I was quite certain that I was the first to recognize that proofs of general computation and blockchain could be very well combined. Ethereum had not yet emerged ten years ago, and when I first met Vitalik, he was not yet the founder of Ethereum but a content author for Bitcoin Magazine. We discussed many issues regarding proofs, validity proofs, and zero-knowledge proofs.

One of my early co-authored papers was titled "Zerocash: Decentralized Anonymous Payments from Bitcoin," which discussed how to use zero-knowledge proofs to protect payments on Bitcoin. This became the later Zcash project, and we have been developing this technology ever since. Around 2018, we proposed better technical improvements and called it STARK, which helps scale Ethereum or general blockchains. In 2018, we co-founded StarkWare, with founders including myself, Uri Kolodny, Alessandro Chiesa, and Michael Riabzev. Today, it is widely believed that our validity proofs are the best way to scale Ethereum.

BlockBeats: You mentioned that you had early contact with Vitalik; I remember he started investing in and supporting StarkWare quite early.

Eli: Yes, Vitalik was one of the early investors in StarkWare. This is one of the few investments he has made to date, and we are very proud of it.

I remember that Vitalik became a core investor around the fall of 2018. I attended a meeting with him, seemingly organized by Tsinghua University, but it was definitely in Shenzhen. I remember we were walking through a very nice seafood market, discussing things like StarkWare, his investments, and a bunch of other topics. So the details of that decision were made in China.

BlockBeats: You started considering using ZK proofs as a scalability solution before Ethereum; how did you foresee that blockchain would inevitably face scalability issues?

Eli: I believe readers in the Chinese community must be familiar with Alipay and WeChat Pay, which may have transaction speeds reaching one million per second. Now, blockchain needs to handle NFTs, games, voting, elections, and governance. To make all of this happen, the current technology is not good enough; on Bitcoin and Ethereum, there are about 10 to 20 transactions per second. This requires other technologies that do not compromise blockchain security but can expand the blockchain's scope or reach, which is what Starknet does.

BlockBeats: Professor Ed Felten, the founder of Arbitrum, also started addressing scalability issues around 2014, but he decisively chose interactive fraud proofs, which are later widely regarded as more feasible optimistic fraud proofs. Why did you decide to use ZK as a solution from the beginning?

Eli: It is a better technology, and it does work. In fact, everyone in the Ethereum core team knows that Optimistic Rollups did not initiate their core technology. There are no fraud proofs on Arbitrum and Optimistic. The reason they haven't launched is that it simply doesn't work.

In contrast, we have never submitted a state update without a STARK validity proof. Even if our system's performance reaches 10 times that of Ethereum, we have used less than 1% of Ethereum's resources. Optimistic Rollups are a good idea, but they do not work, and because they do not work, they have not been initiated. There is a significant difference between something that should work but does not and something that does work and can be proven.

BlockBeats: But although Vitalik invested in StarkWare, he later supported Optimism. There seems to be an increasingly strong consensus in the crypto space that scalability issues should first start with Optimistic and then shift to ZK, because ZKP or ZK Rollup is technically much more difficult to implement than Optimistic. What do you think about this?

Eli: I like the Optimism team; they are outstanding members of the community, but the technology is not good. It's like asking why we can't attach feathers to people and let them flap their wings to build airplanes? That does make more sense, which is why Leonardo da Vinci painted people with wings. But it does not work in mathematics and economics. I really like and admire their intelligence and the truly great members of the community, but that does not change the fact that it is a technology that cannot work well.

BlockBeats: But as an ordinary user, Optimism and Arbitrum are indeed practical solutions for now.

Eli: Of course, but it needs to be recognized that your asset security relies on trusting a single operator, and then it works very effectively. There is another possibility that works well. You can bridge your assets to any computer you trust very much. It could be a centralized exchange or something similar, and then you will get amazing TPS throughput, but you are entrusting all your funds to that computer controlled by that person. If it fails, goes offline, or acts maliciously, then all your funds will disappear.

Arbitrum, Optimism, and Binance share the same security assumptions. Binance is a good product for users, possibly even better than Arbitrum, but it is important for users not to be fooled into thinking they have a higher level of security when transferring assets. If someone says I won't use Binance because it is controlled by a set of centralized processes, I want decentralization, I want to put my assets on Arbitrum or Optimism, but in fact, this is the same thing.

BlockBeats: So when users use Optimism or Arbitrum, they are essentially using a centralized trading platform?

Eli: What they are using is a centralized system, and even worse, it is based on a codebase that was not designed from the beginning to be operated by a single operator. So I think the legal protection it provides you is even less than that of Binance.

BlockBeats: Back to ZKP, you established zk-STARK in addition to the widely recognized zk-SNARK. Why set two different technical standards for ZK technology? And why did StarkWare ultimately choose zk-STARK?

Eli: The SNARKs technology, Groth-16 type of preprocessed SNARKs, uses elliptic curves and can be prepared for production earlier, but I believe it is a lesser technology in terms of efficiency and security. It's like having a technology that is not that good but appears first, similar to how the steam engine is not as good as the gasoline engine, but it appeared first.

As for why we chose zk-STARK, it is a choice that does not require much thought. It has the safest and most future-proof foundational technology, relying on battle-tested mathematics and cryptographic techniques. It is worth mentioning that it is post-quantum secure, which other systems cannot achieve. It has the fastest and most efficient proofs, and because of these, it reaches the highest level among all proof technologies.

Comparison of zk-STARK and zk-SNARK technical details, image source from Horizon Academy

BlockBeats: Currently, many ZK Rollup and zkEVM projects are using zk-SNARK. What are your thoughts on the future of zk-SNARK?

Eli: I believe it will gradually be phased out. It may remain for personal shielded transactions, where users may not care about the inefficiency of proof times. And indeed, there are very short proofs, about 200 bytes. So they may have some utility in the privacy domain for personal shielded transactions, but once it reaches larger throughput, it can only go that far. All projects in the Polygon system are using the STARKs we invented, from Hermez, Miden to Polygon Zero, etc. Some projects are currently temporarily using SNARKs, but ultimately they will all shift to STARKs. So now the largest number of blockchain validity proof projects are using STARKs.

On Starknet, Cairo, and zkEVM

Many people find it difficult to understand that as an Ethereum L2, Starknet itself is not compatible with the Ethereum EVM, and developers need to use a completely new language called Cairo for contract development. This seems to make Starknet lose many advantages of EVM-compatible L1 and Ethereum-equivalent L2, namely to minimize development costs and quickly launch the ecosystem in a short time. However, in Eli's view, these issues become less significant in light of the enormous potential brought by Cairo. More importantly, the Starknet ecosystem already has its own solutions for this.

BlockBeats: Why did StarkWare establish two different projects, StarkEx and Starknet? Why not combine the two?

Eli: StarkEx and Starknet serve two different purposes. When you talk about a very specific type of high-throughput transaction, such as minting and trading NFTs, you can get better performance and throughput from a system specifically built for that purpose, which is StarkEx. We started with a very specific use case and tested the entire technology in practice, which we are very proud of.

Starknet, on the other hand, went through a more complex build first, so it naturally appeared later, and it serves a different purpose, allowing you to perform general computation, just like Ethereum. But this general computation means that its performance will not be as good compared to specific performance designed for a particular use case. So, it is a trade-off between generality and efficiency.

BlockBeats: So for the StarkWare team, achieving high transaction TPS in specific application scenarios is very important?

Eli: Yes, it is very important for us to achieve this goal. Many teams "run their projects on Twitter," they publish some nice numbers and then expect people to be fooled into thinking this is an effective system. But we take pride in our high TPS system that has been running for years. Everything is aimed at increasing throughput and reducing transaction costs.

I want to say that so far, our system has processed nearly $1 trillion in transactions, over 450 million transactions, and issued over 100 million NFTs. No other L2 can achieve this. Our daily TPS is 50% higher than Ethereum, and our weekly TPS is 30% higher than Ethereum, sustained over an entire week. No other system can reach this level, and we have never used more than 1% of Ethereum Gas, even as our TPS increased by 30%.

BlockBeats: StarkEx is a very specific use case. As a technology-driven project, StarkWare seems to have had a very clear positioning for its business model from the beginning.

Eli: I often like to say, "Look up at the stars, and keep your feet on the ground." We understand mathematics, engineering, and technology very well, and it is easy to be attracted to the next cool scientific project and enjoy showcasing some trendy new things, which is very tempting for me. But we set aside the flashy content that cannot meet users' immediate needs and instead focus very intently and diligently on providing what users and customers truly need. As a scientist, the whole team knows that I feel very regretful for not being able to do things like elliptic curve fast Fourier transform (ECFFT) and many peculiar techniques that we know could improve performance by 10 times or even 100 times.

StarkWare team members, image source from StarkWare

BlockBeats: Next, I want to talk about the Cairo language. As an Ethereum L2, Starknet is not compatible with the Ethereum EVM, and developers must use the Cairo language to build applications. This seems to make Starknet lose many advantages of the "EVM-compatible ecosystem." Why choose to build your own language?

Eli: Yes, I want to say a few things about Cairo. First, it is an advanced, next-generation smart contract programming language. It supports some new features from the start, such as smart wallets or account abstraction, which are built into the Cairo language. It is a programming language inspired by Rust, and many developers who have tried Solidity and used it, especially those who know Rust, say they prefer programming in Cairo rather than Solidity. So it is already a very, very good programming language.

The second thing I want to say is that the reason for using a dedicated language to create STARK proofs or general validity proofs is similar to why you want to write smart contracts in Solidity. You need to use Solidity not because it has a better developer experience than Python, Rust, or C++. On the contrary, its developer experience is worse, but when you transition to a new framework with blockchain infrastructure, you must use a different programming language. Solidity is good because it can effectively compile to the Ethereum Virtual Machine. Let me give another example. If you want to write really good graphics using a GPU, you should use CUDA because it is a language that can maximize GPU performance.

The same goes for STARK proofs. With a different virtual machine, there are different constraints about what makes something valid or invalid. If you want to achieve the performance that STARK proofs can achieve, such as putting a million NFTs in a single Ethereum block, you must use the programming language more efficiently. We can mint ten million NFTs in a proof on Immutable X; I cannot imagine any EVM-based or any zkEVM being able to come close to what we can put in a single proof on the blockchain. I can challenge the world's zkEVM; they will not be able to do it. That is why you need a different computing model and a different programming language.

BlockBeats: But many projects in ecosystems that use Rust, like Polkadot, generally give me feedback about operational costs because hiring or training an engineer who knows how to develop contracts in Rust is very expensive. On the other hand, if the team develops projects on Ethereum or on EVM-compatible L1, it is very cheap. What do you think about this situation? Is this a problem for the Starknet ecosystem?

Eli: This could be a problem, but developers are flocking to Cairo and STARK. The number of developers and teams is increasing every day, and I think they see the potential. Many people understand the issue of scalability and ask themselves what the best and most forward-looking solution is, what truly opens up scalability for global demand. I believe many of them have come to the correct conclusion, which is to take the path of Starknet and Cairo 1.0.

There is an analogy: usually, you would write the first version of your software in Python, but when you want to achieve scalability, you need to write it in other very efficient languages like C++, WASM, or Rust. I believe the same thing will happen with Cairo; you might deploy Solidity code to Kakarot, but that is like writing a high-frequency trading engine in Python, which should not be done. You need to write it in another language, and that other language will be Cairo.

BlockBeats: There are also some projects developing zkEVM based on Starknet.

Eli: Yes, there are some very good projects building ZK-EVMs on Starknet. The first is called Warp, made by the NetherMine team, which is excellent. The second is called Kakarot, which is a community that is forming.

Kakarot is a ZK-EVM built on top of Cairo. So I am very confident that most ZK-EVMs will be deployed on Starknet. If you already have the idea of porting your Solidity code or EVM code to L2, the best way is to do it on top of Kakarot or Warp. So this again showcases the power of Cairo and Starknet.

Competition Among ZK Rollups and the Development of Starknet's Ecosystem

BlockBeats also mentioned the competition among ZK Rollups during the interview, as well as the issues of Starknet's own ecosystem development. As the area with the highest technical content and requirements in the entire crypto industry, ordinary users seem to find it difficult to distinguish between zkSync and Starknet. For most people, the reality is that whoever is better at "playing the game" is favored. From this perspective, StarkWare's philosophy of being driven by technological innovation seems to pose a dilemma for Starknet. How does Eli view the competition among ZK Rollups? What expectations does he have for the development of the Starknet ecosystem?

BlockBeats: In the current ZK Rollup track, Starknet and zkSync are absolute leaders, and there seems to be a strong competitive relationship between the two. In your view, what are the differences between Starknet and zkSync, and what are Starknet's advantages?

Eli: You have to ask yourself, where are the best technological steps? Where is the most likely place to gather the most developers and provide the best, safest technology? We demonstrated our capabilities before zkSync; they are good at storytelling, but there is a significant mismatch between the story they tell and the developer experience. I have a suggestion: I encourage developers reading this article to try developing or deploying something on zkSync and then try developing and deploying something on Starknet; you will feel a distinct difference.

Let me tell you a story: someone previously asked on Reddit who could create something capable of 300,000 transactions; we achieved that on the mainnet, with TPS around 3000. A week or two later, the zkSync team came out saying we have 3000 TPS, but this 3000 TPS has never been measured, proven, or demonstrated; however, it is a good story. There are many people in this industry who are willing to tell stories, and there are also those who present facts; we take pride in our ability to present facts.

BlockBeats: Some users have also reported that the gas fees for transferring assets or interacting on ZK Rollups are currently very high. Why is that?

Eli: Currently, due to high demand, Starknet is experiencing congestion. We will release version 12 in two weeks, which will greatly improve TPS and address most of the demand. Then in the next version, when we fully integrate Volition for off-chain data processing, prices will drop significantly, which may happen in a month or two.

BlockBeats: Speaking of ecosystem developers, some have indeed responded that Starknet is more balanced in ecosystem development compared to zkSync.

Eli: That's right; this is the result of a better and deeper technical understanding. We have been working in this field for a long time, deploying content in production environments and refining the system. We know what we are doing. But we are not very good at storytelling and rarely hype things up. We are like Tesla; Tesla does not have a marketing department; our product is our marketing.

Scene from StarkWare Sessions 2023, image source from StarkWare

BlockBeats: But in the Web3 space, teams that understand marketing seem to be more favored. Many technology-driven projects have encountered issues in ecosystem development, haven't they?

Eli: No, I don't think that will happen next. I believe this because Ethereum has developed this way. One day, there will be a very successful app that clearly demonstrates Starknet's capabilities. What that app is, I do not know right now. If I had to guess, I think it is not a product that currently exists on L1 and can be easily copied and pasted to L2.

So I believe it will make everyone see and talk about it because it is a brand new application that cannot be achieved on L1 or even other L2s. This will be a turning point, and developers will flock to it. That is why we are focusing most of our attention here; we came to EDCON to reach more developers and invite them to join our ecosystem. Because among them, there will be some who create wonderful things that I cannot predict.

BlockBeats: For developers preparing to develop in the Starknet ecosystem, do you have any directional advice for them? What can Starknet offer them?

Eli: I cannot answer that question for them; even if I gave an answer, they should not listen to me; they should decide for themselves. I think there are some amazing things to build on Starknet, like advanced Oracle services, Proof of Humanity, social networks, etc. There are many truly amazing things to do; they should explore and find out what they think is missing or what they find most interesting.

Starknet is the next step in the evolution of blockchain, and it is a very important step. If you are a developer, Starknet will be ready for users; we are building it for the end users. I believe we have a very clear path forward, and there are many tools for more developers to build, and there is much work to be done. There are many opportunities for you to stand out and earn a reputation through hard work.

BlockBeats: Finally, do you have any words for the readers?

Eli: If there were a time machine, you might wish to go back to the early days of Ethereum, where building almost anything you wanted would be very successful, like the first prediction market, the first wallet, etc. But the game is not over; there are next-generation opportunities, and I believe the winners of this generation will be Starknet. So now, besides having a time machine, everyone can learn about Starknet; it is time to do so.