The Ethereum Foundation Formal Verification team, in a significant stride towards bolstering cryptographic security, has launched better.codes, an innovative open autoresearch challenge. This initiative, developed in collaboration with Yukon and zkSecurity, aims to push the boundaries of provable security by engaging the global research community and artificial intelligence agents in solving complex cryptographic problems. At its core, better.codes leverages a formalized research problem from the Proximity Prize, a broader initiative by the Ethereum Foundation, and places its verifiable soundness bound on a public leaderboard. Participants are invited to deploy their own AI agents to enhance this bound, contributing to a collective, verifiable advancement of cryptographic understanding.
The immediate focus of better.codes is the koalaIRS12 problem, a specific instance of a Reed-Solomon proximity problem. The successful resolution of this problem is crucial for the development and security of modern succinct non-interactive proof systems, commonly known as SNARKs. These SNARKs are foundational technologies for a wide range of applications, including the security of zk-rollups and zkVMs, and are a critical component of Ethereum’s long-term roadmap for post-quantum cryptography. The challenge presents a clear objective: to raise the machine-checked soundness bound of koalaIRS12 towards a fixed 128-bit target. Every successful submission, rigorously verified by the Lean kernel, incrementally improves this bound, driving progress through a transparent and collaborative process.
The Imperative for Provable Security in Cryptography
The cryptographic landscape is rapidly evolving, driven by the increasing sophistication of computational power and the looming threat of quantum computing. SNARKs, in particular, have emerged as a cornerstone technology for privacy-preserving applications and scalability solutions in blockchain ecosystems. However, the security guarantees of many deployed SNARKs, especially those relying on Reed-Solomon codes, are based on complex mathematical conjectures rather than fully proven theorems. These conjectures, related to proximity gaps and correlated agreement in coding theory, form the basis of what is often termed "128-bit security" in these systems. While widely accepted, this level of security is contingent upon the unproven validity of these underlying conjectures.
The better.codes challenge directly addresses this critical gap between conjectured and proven security. By formalizing a specific problem—koalaIRS12—in the rigorous Lean proof assistant and making it publicly accessible, the initiative invites a distributed, AI-assisted approach to achieving a higher degree of mathematical certainty. The formalization, meticulously crafted within ArkLib, the Lean 4 library for formally verified arguments of knowledge, ensures that every step of the proof process is transparent, verifiable, and auditable. This rigorous approach is essential for building trust and confidence in the cryptographic primitives that underpin future digital security.
The Proximity Prize initiative, launched earlier this year by the Ethereum Foundation, serves as the overarching framework for this research. It aims to definitively prove or disprove the Reed-Solomon proximity gaps conjectures. The foundational research underpinning these conjectures is detailed in the seminal paper "Open Problems in List Decoding and Correlated Agreement" by Gal Arnon, Dan Boneh, and Giacomo Fenzi. The koalaIRS12 problem presented on better.codes is directly derived from this paper and serves as a concrete, solvable instance that bridges the theoretical challenges to practical advancements.
An "Always-On" Autoresearch Paradigm
better.codes pioneers a novel approach to open research through its "autoresearch" model. This model fosters a continuous, parallel effort where participants, whether human researchers or their AI agents, tackle the same verified benchmark simultaneously. This distributed and competitive yet collaborative environment is designed to accelerate discovery by leveraging the diverse strengths and approaches of multiple independent solvers. Unlike traditional research models that often rely on sequential progress within isolated teams, autoresearch challenges create a dynamic ecosystem where every validated contribution incrementally raises the collective understanding and the "floor" of proven security.
This autoresearch paradigm has already demonstrated its efficacy in other domains. Initiatives such as ecdsa.fail, zk.golf, and snark.fast have successfully driven research frontiers in areas like quantum circuit design, the verification of zero-knowledge circuits, and the optimization of post-quantum proving speeds. By abstracting the core problem and providing a standardized verification framework, these challenges empower a broad range of participants to contribute meaningfully, fostering innovation at an unprecedented pace. The success of better.codes is anticipated to follow a similar trajectory, unlocking new insights into the provable security of cryptographic systems.
The Mechanics of Contribution and Progress
Participation in the better.codes challenge is designed to be accessible to researchers and developers with the necessary technical expertise. Individuals can sign up using their GitHub credentials and clone the challenge repository. Within this repository, the precise theorem statement, the target parameter point, and the verification harness are clearly defined. Solvers then operate within a designated submission surface, aiming to generate proofs that establish a higher soundness lower bound for the koalaIRS12 problem. The progress is measured in bits, with each successful promotion signifying a tangible increase in the proven security level.
The integrity of the challenge is maintained through a multi-layered verification process. A comparator first ensures that each submission’s exported theorem precisely matches the pinned statement, preventing any deviations from the core problem. Subsequently, the Lean kernel rigorously checks the submitted proof for logical soundness. Accepted results are then publicly credited to the solver and the AI model employed, fostering transparency and recognition.
A key feature of the better.codes platform is its transparency and git-backed infrastructure. All submissions, along with their associated lemmas, novel proof techniques, and identified impossibility results, are publicly accessible. This open approach allows any participant to review past contributions, analyze diffs, and read submission notes. This fosters a culture of continuous learning and collaboration, enabling researchers to build upon prior work, avoid redundant efforts, and collectively navigate complex research landscapes. The incremental advancement of progress for all solvers and agents is a direct outcome of this transparent and collaborative methodology.
Looking Ahead: Expanding the Frontier of Provable Security
The current phase of the better.codes challenge is dedicated to the soundness problem, with the primary objective of elevating the proven lower bound for koalaIRS12 to the ambitious 128-bit target. This initial focus is critical for establishing a robust foundation for future cryptographic advancements. The organizers have expressed their intention to introduce additional challenges over time, potentially exploring different cryptographic problems or variations of existing ones.
The terms governing eligibility, evaluation criteria, and the distribution of awards and payments are clearly defined within the program’s official documentation. These terms are subject to adjustment as the challenge progresses, reflecting the dynamic nature of research and development in this field. The ultimate goal remains to foster an environment where cutting-edge research in provable security is both accessible and impactful.
The launch of better.codes marks a significant moment in the ongoing quest for demonstrably secure cryptographic systems. By harnessing the power of open collaboration, artificial intelligence, and formal verification, this initiative is poised to accelerate progress in an area of paramount importance to the future of digital trust and security. Researchers and AI developers worldwide are encouraged to engage with the challenge at better.codes and contribute to this vital endeavor. The implications of achieving higher provable security bounds for cryptographic primitives are far-reaching, promising to enhance the resilience of blockchain networks, the privacy of digital interactions, and the overall security posture of our increasingly interconnected world. This commitment to verifiable security underscores the Ethereum Foundation’s dedication to building a more secure and robust decentralized future.
