Interview with Brevis CEO Michael: zkVM Scaling Far More Effective than L2

In a Consumer-Grade Hardware Environment, 64 GPUs Completed 99.6% of the L1 Block Proof in 12 Seconds—Brevis's multi-GPU zero-knowledge virtual machine (zkVM) Pico Prism achieved a significant performance breakthrough.

This milestone progress has also sparked high attention from the Ethereum core community. Ethereum's founder Vitalik Buterin publicly commented, "I'm glad to see Brevis's Pico Prism officially entering the ZK-EVM verification field. This is an important step for ZK-EVM in terms of proof speed and diversity."

"The data itself speaks volumes." Brevis Co-founder and CEO Michael said, "We have built infrastructure capable of processing real-time Ethereum block output using consumer-grade hardware. This performance is the best response to Ethereum's decentralization goal."

Brevis Co-founder and CEO Michael—a UIUC computer science Ph.D., distributed systems, and high-performance networking expert—has successfully started and exited multiple Silicon Valley startups, with his research results being adopted by numerous tech companies, including Google. In 2018, he founded Celer Network (peaking at a valuation exceeding $2 billion). Today, he leads the Brevis team in combining zero-knowledge proofs with verifiable computation, aiming to build infrastructure that can truly support Ethereum's ecosystem's large-scale computational needs.

To better understand how Brevis achieved this breakthrough and what it means for the Ethereum ecosystem, DapTran BlockBeats conducted an in-depth interview with Brevis Co-founder and CEO Michael.

We explored three core questions:

· Brevis's mission and positioning

· The practical value that verifiable computation can bring to applications

· In the process of deep alignment with the Ethereum ecosystem, Brevis's next steps

This is not only a technical discussion but also a forward-thinking consideration of future computing trust.

What Is Brevis: From Vision to Positioning

DapTran BlockBeats: Could you first introduce in the simplest terms what Brevis is as a project and what problem it aims to solve?

Michael: We position Brevis as the Infinite Computing Layer, serving not only Web3 but also broader application scenarios. The core goal is to provide nearly unlimited computing power for on-chain applications while maintaining decentralization and security, addressing the structural contradiction between high performance/complex functionality and trust minimization that has been difficult to reconcile in the long term. Although L2 scaling has significantly increased throughput in recent years, complex computations (such as large-scale historical data processing, cross-chain aggregation, algorithm/AI inference, etc.) are still limited, making it challenging to implement many advanced features.

Brevis's solution is off-chain computation and on-chain validation. The heavy computation is moved to off-chain execution, generating zero-knowledge proofs (ZK Proofs); on-chain contracts only need to validate the correctness of the proof at extremely low cost without reexecuting the entire computation. As a result, contracts essentially receive a set of cryptography-secured plugins, significantly expanding the boundary of usable computing power and functionality without sacrificing decentralization and security. Currently, smart contracts are not very intelligent, but with Brevis, they can truly become intelligent.

BlockBeats: Brevis is referred to as the Web3 Infinite Computing Layer. How did this concept come about? What are the fundamental differences between it and the traditional understanding of on-chain computation or scaling?

Michael: We call Brevis the Web3 Infinite Computing Layer because its goal is to break the computational limit of blockchain. Traditional scaling solutions (such as Layer 2 Rollup) mainly improve transaction throughput from tens of TPS to hundreds or thousands of TPS but still have a ceiling and do not address whether any arbitrary complex computation can be completed on-chain: on-chain usually can only process relatively simple transactions and contract operations. Once the computational complexity increases, it is both difficult to model on-chain and difficult to sustain, such as large-scale data processing, AI inference, or complex algorithms.

The starting point of Brevis is different: We are not just making established types of transactions run faster, but enabling the blockchain to accept and support the results of any computation, while maintaining security and trustlessness. This corresponds to a new paradigm of verifiable computing— as long as off-chain computation can produce a zero-knowledge mathematical proof, the on-chain can confirm its correctness and security at extremely low cost.

For developers, Brevis is more like a cloud of verifiable computation: it seamlessly scales computing power on-demand, offloads complex computations off-chain, submits proofs for on-chain verification, all while maintaining the same security and trust model as native on-chain contracts. Therefore, we define Brevis as an infinite computation layer, achieving nearly infinite computational scalability and availability, while remaining consistent with the trust and security of the blockchain itself.

BlockBeats Interview: Verifiable Computing, as just discussed, sounds quite abstract. Can you explain its role in Brevis more intuitively?

Michael: The fundamental limitation of blockchain is that while ensuring security and trustlessness, computational capacity is limited. Public blockchains like Ethereum belong to consensus systems, where a large number of nodes in the network independently perform the same calculation for a transaction or contract and only produce a new block after reaching consensus. This mechanism is secure but inefficient, similar to the entire class independently working on the same problem to confirm the answer.

The idea of verifiable computing is to decouple computation from verification to reduce redundant costs. Taking the example of the chicken-rabbit cage problem: solving it requires forming equations and deducing, while verifying whether a given answer is correct simply requires substituting the values back into the equation, a far less costly process than re-solving. Computing itself and verifying the computation's result are naturally two different problems in computer theory, with the latter usually being more lightweight. Zero-knowledge proofs extend this to any computation, whether it's a simple operation like 1+1=2 or reasoning of large models. After the computation is completed, a succinct proof is generated, enabling others to quickly confirm at very low cost without revealing the computation details.

In Brevis, heavy computations are performed off-chain and generate a ZK proof, while on-chain, only proof verification is conducted, avoiding the need for repeated execution of the entire computation. This is a core value of verifiable computing, thoroughly separating computation from verification, greatly saving the overhead of numerous recalculations for the same goal.

Why Do This: Technical Architecture and Core Concept

BlockBeats Interview: At a technical level, Brevis consists of two core modules, the ZK Data Coprocessor (referred to as zkCoprocessor below) and Pico zkVM. What is the relationship between these two, and what problems do they each solve?

Michael: You can think of Brevis as a layered structure: the underlying Pico zkVM is a general-purpose computation engine (Virtual Machine) to which you can delegate any computation for execution, generating a zk proof. A key feature is its high modularity, allowing for multiple coprocessors to be attached. The zkCoprocessor is a coprocessor for blockchain data scenarios within this framework, akin to a plugin or add-on. In its current form, the zkCoprocessor acts more like zkVM's memory system, enabling smart contracts to see and understand historical on-chain events (such as user transactions, balances, holdings, etc.).

More specifically, zkVM addresses the ability to verify any computation for a generic problem, while zkCoprocessor is a custom application layer coprocessor designed for blockchain applications. Since smart contracts inherently operate in the present moment and can only access a limited context of the current block, they cannot directly read or compute over a long historical time window or cross-chain state. zkCoprocessor, through zero-knowledge proofs, provides the contract with "a pair of eyes," enabling off-chain retrieval and aggregation of relevant historical data and generating proofs that these data are real and derived from on-chain state. This allows the contract to have memory and memory-based intelligence while maintaining a minimal trust assumption.

BlockBeats Interview: zkCoprocessor enables smart contracts to "see the past," which sounds very groundbreaking. Could you elaborate on how this is achieved?

Michael: Let's take an example of PancakeSwap's historical transaction volume-based dynamic fee discount:

Firstly, zkCoprocessor reads a user's historical transaction records from the blockchain, aggregates the transaction volume based on a rule (e.g., the last 30 days), and generates a zero-knowledge proof for it;

Subsequently, Pico zkVM, upon receiving the validated records, performs further aggregation and computation (e.g., consolidating the trading volume of different pairs into a single metric) and generates a proof of the computation process;

Finally, the result + proof is submitted on-chain, where the contract verifies the correctness of both proofs at once, confirming whether the address has reached the VIP threshold and automatically applying the fee discount for the next settlement period based on this.

Through this method, the contract avoids redundant on-chain reading and computation of massive historical data, gaining the ability to see the past while maintaining trustlessness and cost control.

BlockBeats Interview: You developed your own Pico zkVM instead of adopting the solutions of other zkVM projects. What considerations are behind this decision? How does Brevis differ from other zk projects?

Michael: This is a crucial question. Indeed, there are already many zkVM projects in the market, so why did Brevis choose to build its Pico zkVM from scratch? The core reason is actually quite simple—we believe that existing zkVMs are mostly still in the lab stage and have not been truly designed for large-scale, real-world applications. In terms of overall performance cost and scalability, they are far from ready for large-scale commercial use. In other words, they are more like proofs of concepts rather than systems that can support millions of calls in production.

Brevis faced a very real-world demand scenario from the outset. Top DeFi protocols like PancakeSwap, Euler, Linea, Usual, among others, need to generate millions of zk Proofs every day. If the underlying zkVM's performance does not meet the standard, the entire system would not be able to go live. This forced us to rebuild a zk computation engine truly designed for production environments—this is the Pico zkVM.

Pico has three significant features. The first is ultimate performance. Pico is currently the world's most powerful zkVM. Our latest release, Pico Prism, is already capable of generating real-time proofs for 99.6% of Ethereum mainnet blocks on 64 boards of 5090 (completed within 12 seconds), with 96.8% of blocks even completing within 10 seconds. This means a 3–4x performance improvement and approximately a 50% reduction in hardware costs. In other words, if Ethereum were running on Pico today, its overall verification efficiency could increase by an order of magnitude, marking a true breakthrough in real-time proof.

The second feature is its unique modular architecture. Pico is currently the only zkVM that supports plug-in co-processors. Other zkVMs are usually closed systems that can only perform general computations, while Pico can mount different modules according to different application scenarios. For example, when access to historical data, cross-verification of on-chain records, or running complex financial logic is needed, Pico's co-processor can be plugged in like a plugin to accelerate specific tasks. This allows Pico to have both general and high-performance specialized capabilities, covering a wide range of applications from DeFi in Web3 to AI computing in Web2 with flexibility.

Lastly, it is developer-friendly. We do not want developers to have to be well-versed in cryptography or ZK theory to use Pico. Developers only need to know how to write Rust, enabling them to call directly as if writing a regular program for ZK applications. This significantly reduces the threshold and completely hides the complexity of ZK at the lower level. In summary, we are not an experimental zkVM but a ZK computation engine designed for the real world and real use cases.

BlockBeats Interview: Recently, you announced that Pico Prism achieved 99.6% real-time proof on consumer-grade hardware. Can you talk about the technological breakthrough behind this achievement and what it means for the performance limit of zkVM?

Michael: Let me first explain the importance of this matter. Performance improvement is not merely a technical optimization but is about the future scalability and development of the Ethereum ecosystem. The current Ethereum architecture requires each node to redundantly execute the same computation, which is secure and reliable, but scalability is nearing its limit. The next step requires adopting a new paradigm: having a single node perform extensive calculations and generate ZK Proofs, while other nodes are only responsible for verification. Through this single-point computation, multi-point verification approach, throughput can be increased by an order of magnitude while maintaining decentralization and security, potentially reaching or surpassing Solana's level, all while maintaining the necessary decentralized node scale.

In July of this year, the Ethereum Foundation set a two-year goal: to achieve real-time proof of 99% of blocks using consumer-grade hardware costing less than $100,000. Once this goal is met, almost limitless scaling can be achieved by leveraging the computational power of ordinary hardware. However, up until this point, this has been largely theoretical. Most zkVM solutions still significantly lag behind this standard in terms of coverage, cost, and speed, with many struggling to even reach 90% real-time coverage.

Enter Pico Prism, the first system to truly cross this performance threshold. In our tests, using only 64 NVIDIA RTX 5090 GPUs (costing around $120,000), we achieved real-time proof of 99.6% of Ethereum blocks in 12 seconds, with 96.8% of blocks being provable within 10 seconds and an average proof time of just 6.9 seconds.

Compared to other protocols, Pico has seen a 70% performance improvement and a 50% cost reduction. In other words, it is not only faster but also cheaper, meaning we are now very close to the Ethereum Foundation's target.

BlockBeats: In terms of functionality and goals, does Brevis look more like Ethereum's ZK acceleration layer or a cross-chain verifiable computation cloud?

Michael: In terms of functionality and goals, these two positions are not contradictory. In the short term, Brevis is more like Ethereum and its Layer 2 ZK acceleration layer. We align with the Ethereum Foundation's direction: achieving a 10x to 100x scaling of the mainnet through zero-knowledge proofs; at the same time, real-time proof capabilities will significantly enhance Layer 2 Rollup's interactivity, making cross-chain transactions faster, more efficient, and cheaper, and promoting better integration of state and liquidity.

From a longer-term, more macro perspective, Brevis's positioning is not limited to Ethereum. Our architecture inherently supports multiple chains and we have already partnered with ecosystem collaborators such as BNB Chain (e.g., PancakeSwap) and are advancing collaborations on projects like perpetual contracts on non-Ethereum chains. We also align with Layer 2/multi-chain ecosystems like Arbitrum, StarkWare, and Avalanche. Overall, we aim to develop Brevis into a verifiable computation cloud for all blockchain applications.

Our goal is to make it easy for smart contracts deployed on any chain to use this set of computing services in the next 3–5 years. Looking further ahead, the vast majority of blockchain application computations will occur off-chain and will be guaranteed trustworthy through zero-knowledge methods. Brevis will serve as the unified computing trust layer for the entire decentralized system.

Implementation: Real Applications and Use Cases

BlockBeats Interviewer: Many DeFi applications, such as PancakeSwap, Euler, or Linea, are already using Brevis' technology. Could you provide a specific example of how it works in these scenarios?

Michael: Yes, our current use cases are mainly divided into several categories, with DeFi being the most typical one. As mentioned earlier, PancakeSwap is a standard example. We don't want the DEX user experience to always remain uniform—where the interface and fees are the same for all users, regardless of whether you are a regular trader or a high-frequency trader. We hope that DEXs can gradually provide a differentiated experience for different types of users, similar to CEXs. At the same time, we are also exploring more innovative applications. For example, we collaborate with projects like SocialFi to allow users to prove their influence and real holdings without exposing their main wallet.

We are also advancing some scenarios that operate almost entirely off-chain. For example, in a perpetual trading system like Hyperliquid, users' positions and leverage information are often public and vulnerable to front-running by other traders. We aim to refactor such a system using ZK technology to encrypt and hide all order and position records, while still ensuring matching correctness and system security through ZK proofs. This approach can bring a near-centralized exchange-like smooth experience while maintaining blockchain-level security and privacy.

BlockBeats Interviewer: On the product level, Incentra is a significant application you have launched, which currently manages over $300 million in incentive distribution. How does this system operate?

Michael: The premise of Incentra is straightforward: incentive distribution is at the core of every ecosystem's growth, but historically, the distribution mechanism has often faced issues of security, compliance, and transparency.

The core mechanism of Incentra is for users to generate ZK proofs based on their real on-chain or in-protocol behavior, complete eligibility verification and claiming directly on-chain through smart contracts, without relying on centralized distribution. Compared to traditional practices, this model has advantages in three areas. First, security: funds are settled within the contract according to predefined rules, reducing centralized custody and operational risks. Second, compliance and audibility: both incentive rules and eligibility proofs are traceable on-chain, avoiding compliance pressure to distribute funds to unknown entities. Third, fairness and transparency: any claiming behavior can be independently verified externally for its corresponding real contribution and calculation path, eliminating black-box distribution. Currently, this system has been implemented in the growth and incentive scenarios of projects such as BNB Chain, MetaMask, OpenEden, Usual, and others. Whether it's the stablecoin ecosystem or on-chain incentive systems, they can achieve secure, compliant, and transparent incentive distribution through this solution.

Rhythm BlockBeats: Compared to many ZK projects that are still in the experimental stage, Brevis has already achieved large-scale implementation. Why do you think you were able to do this faster?

Michael: I believe the most core reason is that from day one, we did not see ourselves as a ZK research laboratory, but as an infrastructure company focused on real-world applications. Our technical roadmap was not to build an engine first and then find uses for it, but rather the opposite: starting from application needs, reverse-engineering the design of the underlying architecture.

This actually determines the fundamental difference of Brevis. Many ZK projects first develop behind closed doors a seemingly perfect zkVM, and then think about where it can be applied; whereas we started from the real pain points of our collaborative partners, first understanding their actual needs, then gradually designing a modular zkVM architecture, and further creating an attachable zkCoprocessor.

In other words, Brevis' technological evolution is not based on empty planning, but rather step by step driven by real users and partners. This problem-oriented R&D approach allows us to iteratively improve rapidly, with each enhancement directly corresponding to a verifiable use case. This also enabled our system to have production-level stability and scalability right from the start, rather than coding for papers.

There is also a more down-to-earth reason: our team understands both cryptography and large-scale systems engineering very well. We have never stayed at the theoretical level to do deductions, but have truly honed performance and stability verification in a production environment. So far, Brevis has generated over a hundred million zero-knowledge proofs on the mainnet, served approximately 190,000 users, and managed over $300 million in real incentive funds for collaborative protocols. These are all real online applications—such as projects like Linea, Euler, Usual, OpenEden—not demos, and not testnets.

Where to Next: Open Scenarios and Future Vision

Rhythm BlockBeats: Besides DeFi, in which other directions do you think the off-chain computation and on-chain verification model can have an impact in the future, such as AI, data markets, or gaming?

Michael: Beyond DeFi, this model will continue to have an impact in verifiable AI, Web2 to Web3 data and identity bridging, privacy-preserving data markets, gaming, and social scenarios. First, in the verifiable AI field, existing models are mostly black boxes. Through zero-knowledge proofs, the inference process can generate verifiable mathematical evidence, proving that a specific model produced the corresponding output for a given input, while ensuring that the model and process have not been replaced or tampered with. Secondly, in the bridging between Web2 and Web3 scenarios, users can bring qualifications or behaviors from centralized platforms into on-chain applications in a provable form, for example, mapping the transaction activity on a centralized trading platform to fee discounts on a decentralized trading platform, without revealing identity or underlying account details. Finally, in the gaming and social fields, users can prove their historical achievements, asset holdings, or completion of key behaviors to gain access, eligibility, or incentives, without disclosing wallets or personal sensitive information.

In the long run, this system will redefine the trust boundary of the digital world. We aim to reconstruct the entire human societal understanding of trusted computation through Web3 and zero-knowledge proofs, enabling every computation to be verified and trusted. What Brevis aims to do is to become the foundation of all this—a new computational trust layer.

BlockBeats Interview: We saw that on October 13th, Brevis launched the Brevis Proving Grounds new event. Could you share what the core content of this event is, and what users can truly participate in and experience?

Michael: In fact, over the past period of time, we have devoted a lot of effort to validating and approving applications for real-world use. We have now deep collaborations with over twenty partners, each project being a real-world deployable application that users can directly use, and these projects have shown very good data performance and user retention.

But we found that in the world of blockchain, the drive for many things actually comes from user awareness pushing development forward. That is to say, when users truly understand the value of a new technology, their feedback and demands will in turn drive developers to implement more functionalities. It is for this reason that we want to involve users, letting them not only be observers but individuals who can personally experience, validate, and verify. This is the core purpose of our launch of The Brevis Proving Grounds event.

The entire event is roughly divided into two stages. The first stage is the user education stage, where we hope to have more people understand what our partners are doing and what specific problems zero-knowledge proofs (ZK) can solve in these projects. This stage leans more towards popularization and understanding, making it clear to everyone the logic and significance behind this technology.

The second stage is the actual experience stage, where users are encouraged to use these applications that we have already deployed, truly feel the changes brought about by these products after integrating with Brevis. Unlike many projects, where users are usually just clicking buttons, completing tasks, or earning some rewards because they often lack genuinely experiential, underlying innovative functionalities, we are different. Our partner projects themselves are real-world deployable, valuable applications, and users participating are not engaging in a simulated experience but are actually using products that are deeply integrated with Brevis technology.

Through this approach, we hope to enable users to tangibly experience the seamless user experience and transformative value brought by ZK technology, understanding why this off-chain computation and on-chain validation model can make applications both secure and efficient. More importantly, we want to make more users ecosystem drivers: as more and more users understand and actively demand that the products they use adopt this mechanism, the innovation drive of the entire industry will be invigorated in reverse.

Rhythm BlockBeats: Apart from this event, what are Brevis's current plans? Can users participate in them as well?

Michael: We will continue to roll out participatory ecosystem plans, including the launch of a decentralized proof-of-stake node network to form a more open proof and validation infrastructure. The relevant plans will open participation channels to the community and ecosystem partners, and specific participation methods and schedules will be announced to the public in the future.

Rhythm BlockBeats: Looking at a longer-term perspective, what kind of infrastructure do you hope Brevis will become? What role does it ultimately want to play in the Web3 world?

Michael: Our vision is to build Brevis into an infinite computing layer of Web3, serving as the source of universal trusted computing. Analogous to Ethereum restructuring asset and financial trust logic with smart contracts, Brevis aims to reconstruct trust logic in computing: no longer a choice between self-computation and trusting others' computation results, but rather ensuring the security and correctness of results through verifiable computation without duplicate execution.

Looking ahead, on-chain applications will evolve from simple transaction execution and asset transfers to smarter, more complex systems, covering advanced algorithms, data interaction, personalized experiences, and social proof of behavior. To achieve widespread adoption, it must possess strong computing power and comprehensive security simultaneously. We aim to make this proof-based computation universal, user-friendly, and cost-effective—allowing every Web3 application to naturally embed this capability without reinventing the wheel.

Architecturally, we hope to add a new trust layer on top of the consensus layer, data layer, and execution layer—an infinite computing layer. It will become the blockchain's fourth-layer infrastructure, offering unified computing trustworthiness and verification capabilities to the entire decentralized world. All intelligent, complex, and even cross-chain computations can be verified and used in a trustless manner. This is our understanding of infinite computing and the long-term mission of Brevis.