The Ethereum Foundation Formal Verification team, in a significant collaboration with Yukon and zkSecurity, has officially launched better.codes, an innovative open autoresearch challenge designed to push the boundaries of cryptographic security. This initiative represents a novel approach to accelerating research by leveraging the collective power of AI agents and human ingenuity to tackle complex, formalized mathematical problems. The challenge is now live, inviting researchers and developers worldwide to contribute to advancing the field of succinct non-interactive arguments of knowledge (SNARKs), a critical technology underpinning various blockchain applications and future cryptographic standards.
At its core, better.codes presents a self-contained problem derived from the Proximity Prize research. This problem has been meticulously formalized in Lean, a powerful proof assistant, and its "soundness bound"—a measure of its cryptographic security—is publicly displayed on a leaderboard. Participants are encouraged to deploy their own AI agents, or utilize existing tools, to enhance this soundness bound, striving to reach a formidable 128-bit security target. The challenge specifically focuses on koalaIRS12, a Reed-Solomon proximity problem that holds immense importance for the development of modern SNARKs.
The significance of this challenge lies in its direct impact on the security guarantees of numerous blockchain technologies. Many production hash-based SNARKs, including those essential for securing zk-rollups, zero-knowledge virtual machines (zkVMs), and even Ethereum’s post-quantum cryptographic roadmap, rely heavily on Reed-Solomon codes and the associated proximity gaps and correlated agreement principles. The current security benchmarks for these systems often rest on theoretical conjectures, meaning their full 128-bit security assurance is contingent upon the mathematical validity of these unproven assumptions. better.codes aims to bridge this gap between conjectured and proven security by fostering an environment of open, incremental, verifiable, and public research.
This initiative builds upon the broader Proximity Prize, launched earlier this year by the Ethereum Foundation. The Proximity Prize is a dedicated effort to rigorously prove or disprove the fundamental conjectures surrounding Reed-Solomon proximity gaps. These grand challenges were meticulously laid out in the seminal paper, "Open Problems in List Decoding and Correlated Agreement," authored by Gal Arnon, Dan Boneh, and Giacomo Fenzi. The koalaIRS12 problem at the heart of better.codes is directly drawn from this research and is fully formalized within ArkLib, the Lean 4 library for formally verified arguments of knowledge, ensuring a robust and auditable foundation for all contributions.
The Imperative of Provable Security in Cryptography
The drive behind better.codes is rooted in the critical need for provable security in the rapidly evolving landscape of cryptography. Modern digital systems, especially those handling sensitive data or financial transactions, demand robust security guarantees. SNARKs, in particular, have emerged as a cornerstone technology for enhancing privacy and scalability in blockchain networks. Their ability to provide succinct and verifiable proofs without revealing underlying data makes them indispensable for applications like zero-knowledge rollups, which aim to scale Ethereum, and for future secure communication protocols.
However, the security of many deployed SNARKs is intrinsically linked to the theoretical properties of error-correcting codes, such as Reed-Solomon codes. The security bounds, often targeted at 128 bits to withstand sophisticated attacks, are only fully realized if certain mathematical conjectures hold true. These conjectures relate to the "proximity gaps" in coding theory, which dictate how reliably information can be recovered from noisy or partially corrupted data. If these conjectures are proven false, the security assumptions underpinning these cryptographic systems could be significantly weakened, potentially exposing them to vulnerabilities.
The Proximity Prize and, by extension, the better.codes challenge, are direct responses to this critical research area. By aiming to formally prove or disprove these conjectures, the initiative seeks to solidify the theoretical foundations of SNARKs and provide a higher degree of certainty about their security. The target of 128-bit security is a widely accepted standard in cryptography, designed to be computationally infeasible for even the most powerful adversaries, including those that might emerge with advancements in quantum computing. Achieving this level of proven security is paramount for building trust and ensuring the long-term viability of blockchain technology and other cryptographic applications.
An "Always-On" Autoresearch Model for Collaborative Advancement
better.codes pioneers a new paradigm in open research: the "autoresearch challenge." This model fosters a continuous, collaborative environment where participants are empowered to run their own AI models, computational harnesses, and custom tools in parallel against a shared, verified benchmark. The core principle is that no single algorithmic approach is universally optimal for solving complex research problems. By encouraging a diverse array of independent setups to tackle the same benchmark, the collective effort can accelerate progress at a pace unattainable by any single team or organization.
This innovative approach has already demonstrated its efficacy in other prominent research challenges. Initiatives like ecdsa.fail, zk.golf, and snark.fast have successfully pushed research frontiers in areas such as quantum circuit design, the verification of complex ZK circuits, and the optimization of post-quantum proving speeds. The success of these prior challenges underscores the power of decentralized, competitive, yet collaborative research efforts.
The "always-on" nature of better.codes means that the research process is dynamic and continuous. As participants submit their findings, each successful promotion of a proof on the leaderboard incrementally raises the "floor" of proven security. This creates a virtuous cycle of improvement, where each advancement builds upon previous discoveries, making the overall problem more tractable for subsequent solvers. The transparency of the platform ensures that new lemmas, innovative proof techniques, and even discovered impossibility results are not siloed but are instead "upstreamed" into the public repository. This open sharing allows all participants, both human and AI, to learn from each other’s successes and failures, avoid redundant efforts, and collectively steer research towards the ultimate goal.
The Mechanics of Contribution and Verification
Participating in the better.codes challenge is designed to be accessible yet rigorous. Interested individuals can sign in using their GitHub accounts at better.codes, providing a secure and widely adopted authentication method. Once authenticated, participants can clone the challenge repository, which contains all the necessary components: the precise theorem statement to be proven, the specific parameter point under investigation, and the verification harness that will be used to validate submissions.
Solvers operate within a designated submission surface, where they can develop and test their proof-generating strategies. The objective is to prove a soundness lower bound that is demonstrably larger than the current best, measured in bits. Each submission undergoes a two-tier verification process. First, an automated comparator ensures that the exported theorem from the submitted proof precisely matches the pinned statement within the challenge. This step guarantees that participants are indeed addressing the intended problem. Second, and crucially, the Lean proof assistant’s kernel meticulously checks the validity of the submitted proof itself. This formal verification process eliminates ambiguity and ensures that only mathematically sound proofs are accepted.
Accepted results are then promoted to the public repository, a testament to the solver’s contribution. Each successful submission is credited to the solver and, importantly, to the specific AI model or agent employed, fostering transparency and encouraging further development in AI-assisted theorem proving. The entire submission process is transparent and git-backed, meaning every change, every proof, and every piece of discovered knowledge is recorded and auditable. This version-controlled approach allows anyone to examine past diffs, read submission notes, and understand the evolutionary path of the research. This incremental, collaborative approach ensures that progress is built upon a shared foundation, enabling participants to learn from prior work, bypass identified dead ends, and collectively propel the state of the art forward.
Looking Ahead: Expanding the Frontier of Cryptographic Research
The initial launch of better.codes focuses on the soundness challenge for koalaIRS12, with the immediate goal of raising the proven lower bound to the target of 128 bits. This objective represents a significant milestone in formally verifying the security assumptions underpinning critical cryptographic primitives. However, the platform is envisioned to evolve and expand over time. The organizers intend to introduce further challenges that explore different facets of cryptographic research, potentially delving into other areas of coding theory, formal verification of complex algorithms, or novel cryptographic constructions.
The eligibility criteria for participation, the methodologies for evaluating contributions, and the framework for awards and payments are all governed by the program’s terms and conditions. These terms are subject to adjustment as the challenge progresses, reflecting the dynamic nature of research and development. The Ethereum Foundation and its collaborators are committed to fostering a vibrant and productive research community, and the program’s structure is designed to adapt to emerging needs and opportunities.
The ultimate aim of better.codes extends beyond a single challenge. It seeks to establish a sustainable model for open, collaborative autoresearch that can be applied to a wide range of complex scientific and technological problems. By democratizing access to cutting-edge research challenges and empowering both humans and AI to contribute, the initiative promises to accelerate discovery and innovation in critical fields like cryptography, formal methods, and artificial intelligence.
The call to action is clear: individuals and organizations interested in contributing to the future of cryptographic security are invited to visit better.codes and begin their participation. By engaging with this challenge, they can not only advance their own understanding and capabilities but also play a direct role in building a more secure and trustworthy digital future, underpinned by formally verified cryptographic guarantees. The success of this initiative could have far-reaching implications, influencing the design of future blockchain protocols, secure communication systems, and the overall landscape of post-quantum cryptography.
