An innovative open autoresearch challenge, dubbed better.codes, has officially launched, aiming to significantly advance the field of provable security within cryptography. Developed through a collaborative effort between the Ethereum Foundation’s Formal Verification team, in partnership with Yukon and zkSecurity, this initiative invites the global research community to push the boundaries of machine-checked security guarantees for modern cryptographic systems.
At its core, better.codes leverages a self-contained problem meticulously formalized in the Lean proof assistant, originating from the Proximity Prize research initiative. This problem’s soundness bound is now publicly displayed on a leaderboard, creating a dynamic environment where participants can collectively strive to elevate its proven security level. Solvers are encouraged to deploy their own artificial intelligence agents to enhance the machine-checked soundness bound of a specific Reed-Solomon proximity problem, known as koalaIRS12. The ultimate goal is to contribute to the advancement of succinct non-interactive proof systems, commonly referred to as SNARKs, which are foundational to many emerging cryptographic applications.
The Criticality of Provable Bits in Modern Cryptography
The launch of better.codes arrives at a pivotal moment for cryptographic research and deployment. A vast majority of production-grade hash-based SNARKs, which underpin critical technologies such as zk-rollups and zero-knowledge virtual machines (zkVMs), as well as playing a vital role in the Ethereum network’s long-term security roadmap against quantum computing threats, fundamentally rely on proximity gaps and correlated agreement within Reed-Solomon codes. These mathematical underpinnings are what provide the security guarantees for these advanced systems.
However, the current state of provable security for these cryptographic constructs often falls short of the theoretical benchmarks that researchers believe are achievable. While deployed systems are designed to offer 128-bit security, this guarantee is only fully realized if the underlying mathematical conjectures hold true. The better.codes autoresearch challenge is specifically designed to bridge this gap between conjectured security levels and rigorously proven security levels. It achieves this through an open, incremental, verifiable, and transparent research process, fostering a collaborative environment for discovery.
Earlier this year, the Ethereum Foundation initiated the Proximity Prize, a significant research endeavor aimed at definitively proving or disproving the Reed-Solomon proximity gaps conjectures. These grand challenges were further elaborated upon in the influential paper "Open Problems in List Decoding and Correlated Agreement" by Gal Arnon, Dan Boneh, and Giacomo Fenzi. The koalaIRS12 problem, at the heart of the better.codes challenge, is directly derived from this seminal work and has been fully formalized within ArkLib, the Lean 4 library dedicated to formally verified arguments of knowledge. This formalization ensures that every submission can be rigorously verified by the Lean kernel, a testament to the precision and rigor of the challenge.
An "Always-On" Autoresearch Paradigm
better.codes embodies a novel approach to open collaboration through its "always-on" autoresearch challenge model. This paradigm empowers participants to run their own AI models, computational harnesses, and custom tools in parallel against a shared, verified benchmark. Each successful submission that advances the benchmark contributes to a collective raising of the research floor, ensuring that progress is continuous and inclusive.
The rationale behind this decentralized approach is rooted in the understanding that no single AI agent or setup can be universally optimal for tackling complex open problems. By encouraging a multitude of independent setups to work on the same benchmark, the frontier of research can be pushed forward more rapidly and effectively than any single team could achieve alone. This model has already demonstrated its efficacy in other significant research endeavors, including ecdsa.fail, zk.golf, and snark.fast. These past challenges have demonstrably advanced research frontiers in areas such as quantum circuit design, the verification of ZK circuits, and optimizing the speed of post-quantum proving systems, underscoring the power of distributed, verifiable research.
Mechanism of the Challenge: How it Works
Participation in the better.codes challenge is designed to be accessible yet rigorous. Individuals can sign in using their GitHub credentials directly on the better.codes platform and then clone the challenge repository. Within this repository, participants will find the pinned theorem statement, the specific parameter point under scrutiny, and the verification harness necessary to guide their efforts. Solvers operate within a designated submission surface, tasked with proving a larger soundness lower bound for the koalaIRS12 problem. The success of each submission is measured and scored in bits, reflecting the strength of the proven security guarantee.
A sophisticated comparator is in place to meticulously verify that each submission’s exported theorem precisely matches the pinned statement. Crucially, the Lean kernel then performs an exhaustive check of the submitted proof. Only those results that successfully pass these stringent verification steps are promoted to the public repository. Each accepted submission is credited to the solver and the specific AI model or approach employed, fostering transparency and acknowledging contributions.
The entire submission process is transparent and git-backed, providing an immutable audit trail of all advancements. Furthermore, new lemmas, innovative proof techniques, and discovered impossibility results are systematically upstreamed to the main repository. This ensures that all participants can readily access and learn from past diffs and submission notes, enabling them to build upon prior work, avoid redundant efforts, and bypass identified dead ends. This continuous cycle of contribution and integration serves to incrementally advance progress for the entire community of solvers and their AI agents.
The Road Ahead: Expanding the Frontier of Cryptographic Security
The initial launch of better.codes focuses on the critical soundness challenge, with the immediate objective of raising the proven lower bound for koalaIRS12 to the coveted 128-bit security level. Looking forward, the platform anticipates the introduction of additional challenges over time, addressing other crucial areas within cryptographic research.
The eligibility criteria for participation, the methods of evaluation, and the structure of awards and payments are all governed by the program’s official terms. These terms are subject to adjustment as the challenge progresses, ensuring adaptability and responsiveness to the evolving research landscape.
The implications of better.codes extend far beyond the immediate goal of solving koalaIRS12. By establishing a framework for verifiable, collaborative, and AI-driven cryptographic research, the initiative has the potential to accelerate the development of more secure and efficient cryptographic primitives. This is particularly vital in an era where the security of digital systems is paramount, and threats, including those posed by advancements in computing power, are constantly evolving. The open nature of the challenge ensures that breakthroughs are shared, leading to a more robust and resilient cryptographic ecosystem for all.
The initiative represents a significant step forward in how complex theoretical problems in computer science and mathematics are tackled. It combines the rigor of formal verification with the power of artificial intelligence and the collaborative spirit of open-source development. The success of better.codes could serve as a blueprint for addressing other fundamental challenges in cryptography and beyond, fostering a future where security is not just conjectured, but demonstrably proven.
The call to action is clear: researchers, developers, and AI enthusiasts are invited to join this groundbreaking effort. The journey to a more secure digital future begins at better.codes, where collective intelligence and computational power converge to forge provable cryptographic guarantees.
