Interview with NEBRA Founder: Reducing Ethereum L2 Scaling Costs through Zero-Knowledge Proof Aggregation | ETHcc2024 Rollup Day Special

2024.07.10

Interview with NEBRA Founder: Reducing Ethereum L2 Scaling Costs through Zero-Knowledge Proof Aggregation | ETHcc2024 Rollup Day Special

"Cryptography above all, no redundant coordination needed." — Shumo, Founder and CEO of NEBRA

2024.07.10 - 09:27:27

Navigating Web3 tides with focused insights

"Cryptography above all, no redundant coordination needed." — Shumo, Founder and CEO of NEBRA

Author: Sunny, TechFlow

Guest: Shumo, Founder and CEO of NEBRA

Scaling Ethereum is one of the major advancements in the Web3 industry in 2024. Typically, scaling solutions fall into two categories: L1 scaling and L2 scaling, each demonstrating advantages in different development domains.

L2 scaling can adopt either optimistic (op) rollups or zk-rollups. Op rollups assume transactions are valid by default and only verify them when challenged, whereas zk-rollups use zero-knowledge proofs to validate all transactions off-chain before submitting them to the main chain. In this space, Base leads in op rollup adoption, while Scroll is among the front-runners in zk-rollups.

Earlier this year, Ethereum founder Vitalik Buterin tweeted his strong expectation that zk-rollups will become Ethereum’s ultimate scaling solution over the next decade.

To further clarify the definition and mechanics of zk-rollups, it's important to highlight their components: the data availability layer (DA), the settlement layer, and their corresponding proofs—state proofs and consensus proofs. Since zk-rollups require validation of all transaction data, both the state transitions and final states of transactions must be stored on-chain, giving rise to these two layers.

Before reaching this "endgame," Vitalik emphasized: "Getting there will require a lot of infrastructure and optimized provers..."

Therefore, today’s discussion aims to gain insights from NEBRA, an L2 scaling startup, to understand the problems NEBRA is tackling and how it integrates zk-rollup infrastructure to achieve this “endgame.”

Currently, generating a zk proof costs $50—a far cry from the endgame. Hence, we invited Shumo, founder of NEBRA, to discuss how he plans to solve the cost issue through zero-knowledge proof aggregation.

This interview dives deep into Shumo’s journey founding NEBRA, the decentralized principles he has cultivated over years of industry experience, and how NEBRA differentiates itself from competing solutions like EigenLayer and Polygon’s aggregation layer through its focus on cryptography and mathematical expertise.

Key Takeaways

The ultimate value of blockchain lies in enabling coordination among people across different countries, eliminating barriers between individuals. A fundamental insight is that coordination is inherently costly.
Scaling blockchains via zk-rollups boils down to two primary functions: data availability (or publishing) and settlement.

In cases requiring state proofs, data availability is critical for verifying state transitions—it ensures that the data needed to validate state changes is always accessible.
In cases requiring consensus proofs, settlement involves reaching agreement on the data state after transactions occur.
Zero-knowledge proofs and verifiable computing allow us to verify the correctness of state transitions without re-executing expensive state transition functions.

We firmly believe that zero-knowledge-based settlement will become the mainstream approach within the next decade.
We regard settlement proofs as one of the most pressing challenges in blockchain, second only to solving data availability (DA). Our strategy includes recursive generation of zero-knowledge proofs—whether Fourier proofs or aggregating thousands of proofs. This allows all such proofs to be merged into a single submission on Ethereum. As a result, settlement proof costs could drop tenfold or more. This is at the core of what NEBRA is developing.
Unlike platforms that may rely on economic games (like EigenLayer), our approach is based purely on mathematics and cryptography.
While EigenLayer provides significant value, avoiding coordination wherever possible and relying entirely on cryptographic solutions—without introducing unnecessary trust assumptions—is also a crucial strategy.
NEBRA’s philosophy is to leverage Ethereum’s existing mature infrastructure—the result of years of community effort and development—to enhance security and functionality, rather than creating redundant coordination layers.
Launching your own Layer 1 chain primarily offers the advantage of capturing sequencer revenue or miner extractable value (MEV), directly increasing your protocol’s income.

Philosophical Foundations Behind Founding NEBRA

TechFlow: Could you briefly introduce yourself? From your academic background to joining Algorand, and now focusing on scaling Ethereum’s infrastructure—what has your journey been like?

Shumo Chu:

I earned my Ph.D. at the University of Washington, where my research focused on formal methods and database systems. This involved theoretical work in programming languages and practical applications in distributed and database systems. My initial interest in blockchain began with reading the Bitcoin whitepaper, and later I even organized a seminar on Bitcoin and cryptocurrencies at the University of Washington back in 2018.

After completing my doctorate, although I wasn’t certain about my career path, I knew I didn’t want to join big tech companies like Microsoft, Google, or Facebook. So I chose to join Algorand—an early cryptocurrency startup with a strong academic reputation. Algorand was founded by Silvio Micali, a Turing Award winner and MIT professor. While the team was excellent, I quickly realized the company culture didn’t align with the grassroots, slightly rebellious spirit I associate with cryptocurrency. Algorand’s decision to partner with large banks diverged from my understanding of crypto’s ethos.

After working at Algorand for a while, I returned to academia as a teaching assistant professor at UC Santa Barbara, shifting my research focus to cryptography, particularly zero-knowledge machine learning (ZKML). I co-authored one of the early papers on ZKML, marking a new chapter in my academic journey. However, my passion for entrepreneurship led me to co-found Manta Network, initially a privacy-focused blockchain project that evolved into a more user-friendly tool known as Manta.

Last July, I left Manta to found NEBRA, a research-driven organization dedicated to advancing the future of zero-knowledge technology. This journey—from academia to entrepreneurship—sums up my path in blockchain and cryptography.

On the Passion for Decentralized, Grassroots Movements

TechFlow: You mentioned anarchism and felt Algorand lacked the grassroots social spirit seen in Ethereum. Could you elaborate on why Ethereum better embodies the spirit of a social movement? Is it simply due to stronger network effects compared to Algorand?

Shumo Chu:

Our projects aren’t just technical—they’re about building communities and realizing founders’ visions. Look at Vitalik Buterin: though he deeply studied cryptography in university and worked with Canada’s top cryptographers, he was ultimately captivated by Bitcoin. He didn’t just observe from afar—he actively participated in the Bitcoin community, serving as editor of Bitcoin Magazine and investing tremendous personal effort. When he realized Bitcoin couldn’t fully realize his vision, he created Ethereum.

This highlights a key principle: cryptocurrency isn’t about maintaining the status quo. It represents a profound affirmation of individual freedom, challenging traditional monetary creation and government intervention. In my career, I’ve witnessed various blockchain trends, including IBM’s Hyperledger. But permissioned blockchains like these have largely faded into obscurity, underscoring a vital point: blockchain technology is inherently anti-central control. It represents a social movement advocating for individual freedom in the face of government oversight.

Blockchain founders like Vitalik openly discuss this dimension, emphasizing that blockchain is rooted in cypherpunk ideology and serves as a tool for social transformation. While central bank digital currencies (CBDCs) are technically feasible, they miss the fundamental spirit of blockchain as a counterbalance to governmental power.

It is precisely this misunderstanding that prevents Algorand and similar projects from truly capturing the essence of the crypto movement.

Building the Ethereum Ecosystem: Coordination as the Ultimate Goal

TechFlow: Before diving into NEBRA, you mentioned the goal of building a peer-to-peer cash system resistant to governments, suggesting a future with governance but not government control. Ethereum is often praised for its smart contracts enabling decentralized applications, while Bitcoin is seen as “digital gold” due to its decentralization, albeit with slower transactions. Given Ethereum’s role in shaping the next generation of internet and finance, could you share your views on these two systems? Also, despite many available networks, why did you choose to focus on Ethereum after leaving Algorand?

Shumo Chu:

I agree with your points. Indeed, Bitcoin serves as a store of value, but it lacks the ability to build a coordination system. The true value of blockchain lies in its ability to help people around the world collaborate, removing friction between individuals across nations. To achieve this, a robust smart contract platform is essential—mere token transfers aren’t enough. Ethereum excels here, embodying the industry’s most fundamental value: decentralization.

While Bitcoin may surpass Ethereum in decentralization, practically speaking, Ethereum remains the preferred platform for building meaningful projects. Its combination of decentralization and widespread adoption is unmatched by any other blockchain.

We shouldn’t focus solely on technical metrics like transaction speed or network latency. Ethereum prioritizes decentralization, which aligns with the industry’s core values. Moreover, the community is actively improving Ethereum through Layer 2 solutions and exploring ways to scale the system without altering the base protocol—for example, using modular data layers like Celestia.

At NEBRA, our primary focus is developing the settlement layer. Fundamentally, blockchain technology revolves around two functions: data publishing (data availability) and settlement. Enhancing Ethereum’s settlement capability aligns perfectly with our goals and is a natural extension of our involvement.

What is NEBRA? How Does NEBRA Scale Ethereum?

NEBRA: Distinguishing Settlement from Data Availability

TechFlow: Can you explain the difference between “settlement” and “data publishing” in Ethereum? It seems data publishing might involve temporarily storing data, while settlement confirms and records transaction states. What distinguishes these processes, and why are they important in the Ethereum ecosystem?

Shumo Chu:

Indeed, you’ve grasped the core concepts—no further explanation is needed. Data availability is crucial for proving state changes—it ensures that the data required to validate state transitions is always accessible.

On the other hand, settlement focuses on reaching consensus about the data state after transactions occur.

Imagine starting from a specific state “A,” where smart contracts define how the state should evolve. The key question becomes: How do we verify a transaction’s correctness? The ultimate solution lies in zero-knowledge proofs and verifiable computing technologies. These allow us to prove the correctness of state transitions without re-executing the computationally expensive state transition functions.

Think of Ethereum’s computational capacity as the calculator you used in fifth grade. This analogy reveals the limitations we’re dealing with. Our goal is to scale Ethereum in a way that preserves its decentralization and trustless nature while maintaining underlying security. This is the central focus of our current efforts.

TechFlow: How does NEBRA balance decentralization, scalability, and security?

Shumo Chu:

The core of our approach is zero-knowledge proofs. We don’t need to re-execute state conditions to verify them. To understand the significance of NEBRA’s method, first grasp concepts like ZK-EVMs and ZK Layer 2 solutions—they use Ethereum as a settlement layer for zero-knowledge proofs. We firmly believe zero-knowledge-based settlement will become mainstream over the next decade, primarily due to its numerous advantages. It maintains security without sacrificing decentralization and offers privacy protection. For instance, one of our key clients, Worldcoin, uses our technology to provide identity privacy for its users.

We believe that after resolving data availability issues, settlement proofs will be among blockchain’s biggest challenges. Our strategy includes recursively generating zero-knowledge proofs—whether Fourier proofs or aggregating thousands of proofs. This enables consolidating all proofs into a single submission on Ethereum. As a result, settlement proof costs could decrease tenfold or more. This is precisely the core of NEBRA’s R&D.

Recursive Zero-Knowledge Proofs and Proof Aggregation: Implications for ZK-EVMs and ZK Layer 2s

TechFlow: Could you simplify the concept of “proof aggregation” for non-technical audiences? Also, please clarify which layer NEBRA operates on.

Shumo Chu:

The precise technical definition of “proof aggregation” isn’t the key point. What matters is understanding what the proof is: it can be a concise signature or a small piece of encrypted data verifying the validity of a state transition function. The benefit is its brevity. However, verifying a proof remains expensive—currently costing up to $50 on Ethereum. At NEBRA, our goal is to reduce this cost to $5 in the short term, and eventually down to 5 cents. Cost reduction is our core value proposition and the primary reason our service is so critical.

As we reduce costs, we do not compromise on trust requirements. Unlike platforms that may rely on economic games (such as EigenLayer), our approach is entirely based on mathematics and cryptography.

Is EigenLayer a Competitor to NEBRA?

TechFlow: Do you consider EigenLayer a competitor to NEBRA?

Shumo Chu:

Our relationship with other teams in this space isn’t competitive; it’s more about different approaches to scaling Ethereum and helping developers build protocols. This topic is nuanced. I don’t see us as competitors. In fact, there are areas where collaboration is possible. While these teams aim to scale Ethereum by adding extra economic assumptions, the effectiveness of these strategies remains to be seen.

At NEBRA, our approach is unique. We aim to leverage Ethereum’s native security features so developers can achieve more without additional constraints. We believe this strategy is the most sustainable in the long run.

TechFlow: Given your deep background in mathematics and zero-knowledge proof research, and EigenLayer’s focus on settlement coordination, how do you envision your development paths converging in the future?

Shumo Chu:

Discussing how to scale and enhance blockchain technology is complex. A fundamental insight is that coordination itself is costly. Sometimes, despite high costs, consensus mechanisms, economic security, or other coordination methods are necessary. Blockchain’s appeal lies in enabling trustless transactions, reducing the friction typically associated with coordination. For example, when transferring Bitcoin, you don’t worry about government stability; Bitcoin remains resilient even without government support, showcasing the technology’s profound advantages.

EigenLayer is building a coordination layer, which I believe holds great value. However, we must acknowledge that not all problems can be solved solely with zero-knowledge cryptography—for instance, double-spending.

Ideally, in scenarios where zero-knowledge cryptography applies, we should minimize reliance on coordination mechanisms. Coordination is expensive. Take EigenLayer’s restaking mechanism: participants restake ETH to earn yields. But we must ask: where do these yields come from? They originate from the revenue of protocols deployed on EigenLayer. As staked amounts grow into billions, required yields increase—if unsustainable, participants may withdraw. Therefore, if a protocol must use coordination, it should do so only when absolutely necessary.

If more efficient technologies like zero-knowledge proofs are available, they should be prioritized. This discussion runs deep. While EigenLayer delivers immense value, the strategy of minimizing coordination and relying entirely on cryptographic solutions is equally important.

TechFlow: Now I understand—we likely need both coordination and cryptographic proofs. Can you elaborate?

Shumo Chu:

Systems should avoid using coordination whenever possible; we should strive not to introduce additional coordination layers.

NEBRA’s Broad User Base: ZK Rollups, ZK Applications, Data Availability Layers, ZK Coprocessors

TechFlow: Could you specify scenarios where zero-knowledge proofs are used, who NEBRA’s potential customers are, and how they can simplify operations without coordination?

Shumo Chu:

To put it directly, our potential customer base is very broad. Specifically:

First, ZK rollups. We can significantly reduce their proof storage costs—by roughly an order of magnitude.

Second, ZK applications, such as Worldcoin, especially those focused on privacy-preserving digital identity, will benefit greatly.

Third, various infrastructure projects, such as data availability (DA) layers. These frequently need to publish proofs on Ethereum, and we can help reduce their associated costs.

Fourth, so-called ZK coprocessors. Among our current partners are ZK coprocessors like Lagrange and Brabus. In the future, ZK coprocessors and ZK virtual machines may see broader adoption.

We see ourselves as a general-purpose protocol suitable for anyone wanting to settle on a chain based on ZK proofs, making NEBRA an even more ideal choice.

We expect continuous optimization of on-chain operations over the next three to five years.

TechFlow: NEBRA aims to use ZK proofs to create a universal settlement layer. But you noted EigenLayer applies in scenarios requiring coordination, where such costs are justified. Can you clarify when a project might need only ZK proofs without EigenLayer, and when both might be necessary?

Shumo Chu:

This touches on nuances in computer science, particularly the concept of “primitives” in computing frameworks. The fundamental principle is: if a process can be fully described in mathematical language, zero-knowledge proofs apply. However, certain problems, like double-spending, fundamentally require consensus mechanisms—a form of coordination.

We need to distinguish specific coordination mechanisms like EigenLayer’s approach from other consensus methods. For example, shared sequencing protocols like Espresso or Astra, or custom consensus algorithms developed by Celestia, serve as alternatives to EigenLayer. When discussing data availability, EigenLayer offers one solution, while platforms like Celestia pursue their own consensus-based strategies.

The discussion shouldn’t be reduced to NEBRA vs. EigenLayer. Instead, it should center on when coordination is needed and when it can be avoided. Coordination via consensus mechanisms is crucial for preventing double-spending, ensuring strong evidence guarantees, and maintaining censorship resistance. Censorship resistance itself is subjective, involving whether information can be included or excluded.

Given that Ethereum already provides a robust consensus layer, I advocate leveraging Ethereum for coordination tasks. By layering ZK proofs atop Ethereum, we can build more efficient systems. NEBRA’s philosophy is to leverage Ethereum’s existing infrastructure—built over years of community effort and development—to enhance security and functionality, rather than creating redundant coordination layers. Our goal is to innovate within Ethereum’s existing framework, using ZK proofs to build more sophisticated systems without additional coordination mechanisms.

Challenges of Zero-Knowledge Proofs as a Scaling Solution

TechFlow: What are the main challenges in reducing the settlement cost of zero-knowledge proofs by tenfold or even hundredfold in the future?

Shumo Chu:

There are many challenges in cost reduction, but they can be broadly categorized:

First, cryptographic challenges. Zero-knowledge proofs are relatively expensive. Although we’ve made tremendous progress over the past decade, ZK development is still in its early stages. There’s much room for improvement.
Second, from the perspective of application developers and users, adopting ZK technology is more complex than not using it. You can see platforms like Arbitrum and Optimism built and deployed optimistic rollups faster than ZK rollup teams. Still, I firmly believe ZK adoption will accelerate, boosting development speed. I believe in the future, building with ZK will be simpler than using traditional techniques, despite the inherent difficulty in designing our proof systems. This is the second major challenge.
The third challenge is market education. Since we’re still early in the ZK space, many don’t yet understand its benefits. For example, users may initially care little about underlying tech. Educating developers about ZK—why it exists and why they should use it to deliver more scalable solutions for Ethereum and other blockchains—is critical. These are the three main challenges I see.

TechFlow: Given research, adoption, and engineering challenges, how are you addressing them simultaneously? Beyond competitors, you emphasize NEBRA’s focus on pure technology and math, while others focus on consensus. Which entities in the industry are working similarly to NEBRA?

Shumo Chu:

We don’t have direct competitors, but the closest to our work is Polygon’s Aggregation (AG) Layer. Both of us are refining aggregation technologies, sharing similarities in using mathematical proof aggregation to scale blockchain, though we approach it differently. Still, key differences exist.

First, Polygon’s AG Layer primarily serves the Polygon ecosystem, but ZK technology’s applications extend far beyond. Outside Polygon’s ecosystem, many ZK rollups and ZK VMs exist—like Score, Caseync, Stackware, Zero, and Sync SP. Our goal is to create a unified layer, potentially even collaborating with Polygon’s AG Layer. Our neutrality and independence from Polygon allow us to work with a wider range of partners. This is a key distinction.

Second, our scope differs. Polygon’s AG Layer focuses on ZK rollups, while we extend beyond to ZK applications. We already have major ZK apps as clients, like Worldcoin, showing our broader applicability compared to Polygon’s AG Layer.

While there are minor differences in technical and performance metrics like speed and universality, these distinctions may diminish over time. Fundamentally, our broader scope and neutral stance set us apart.

TechFlow: How does your multi-layer zero-knowledge settlement solution compare to existing cross-chain bridge solutions?

Shumo Chu:

The core issue is that existing bridge solutions rely on insecure cryptographic primitives, like multisigs, which essentially hand over all your funds to five people you know nothing about. If they collude, you could lose everything. Thus, the fundamental problem with bridges is their highly questionable security assumptions.

Long-term, our goal is to introduce trustless bridging technology to the industry. But that’s not our current focus. Right now, we’re primarily focused on settlement. Bridging is just one method of fund transfer, while our goal is costless interoperability and effective reduction of settlement costs.

NEBRA and Roll-up-as-a-Service

TechFlow: As more applications launch their own chains, you—as a scaling expert—must have unique insights. Could you share the benefits and considerations of launching a dedicated chain for an application?

Shumo Chu:

That’s an excellent question—I’m happy to share my thoughts.

When considering whether to launch a blockchain for your application, understanding the trade-offs is critical. The main advantage of launching your own chain is capturing sequencer revenue or miner extractable value (MEV), directly boosting your protocol’s income. For example, if I launch a DeFi app on my own blockchain, all MEV becomes protocol revenue. This income can be distributed to founding teams, token holders, etc., significantly driving project growth. It’s a straightforward choice.

On the other hand, launching an app-specific chain improves composability, greatly enhancing user experience. If you’re not on your own chain, users typically need to use bridges to interact with other apps or use DeFi components—something I’ve expressed concern about due to security risks, which ultimately degrade the user experience.

This is a fundamental trade-off. But with the emergence of shared settlement layers like NEBRA, the situation may improve, making bridging smoother. While it may not be as seamless as operating on a single chain, it represents significant progress.

Previously, launching a blockchain was too complex, so most opted to build apps. But with services like AltLayer, Gelato, Kadena, and Conduit simplifying the process, we may see more choosing to launch their own chains. By year-end, I wouldn’t be surprised to see thousands of ZK rollups emerge.

As NEBRA, our mission is to support this shift by making chain launches easier and more cost-effective, aiming to enhance interoperability and efficiency.

Additional Discussion on Scaling Solutions: Is Sharding Obsolete?

TechFlow: You mentioned that with the rise of many Layer 2 solutions, Ethereum might face an existential crisis affecting composability, and some prefer sharding over Layer 2. What are your thoughts?

Shumo Chu:

The claim that the blockchain ecosystem is becoming overly fragmented due to multiple Layer 2 solutions isn’t entirely accurate, though there’s some truth to it. Indeed, the ecosystem becomes more diverse with more Layer 2s. But through technologies like NEBRA, we can achieve near-equivalent interoperability in the long run.

A key debate concerns sharding’s effectiveness. Early on, figures like Vitalik Buterin and Carl Beekhuizen (now co-founder of Optimism) proposed Plasma, a sharding scheme. Despite high hopes, Plasma failed to deliver, prompting the community to seek alternative scaling solutions—especially rollups. While rollups have challenges, I firmly believe they’re the right direction. We should focus on developing better protocols to integrate Layer 2 solutions, rather than redesigning sharding mechanisms that previously failed.

Take Polkadot’s sharding approach as an example. Observing Polkadot’s current state offers valuable insights. While criticism of Layer 2s may hold some merit, outright rejection overlooks a reality: we’re unlikely to find more effective alternatives.

Interview with NEBRA Founder: Reducing Ethereum L2 Scaling Costs through Zero-Knowledge Proof Aggregation | ETHcc2024 Rollup Day Special

TechFlow Selected TechFlow Selected