The Ethereum Foundation Formal Verification team, in collaboration with Yukon and zkSecurity, has officially launched better.codes, an innovative open autoresearch challenge designed to push the boundaries of cryptographic security, particularly in the realm of Succinct Non-Interactive Arguments of Knowledge (SNARKs). This initiative aims to systematically enhance the provable security of Reed-Solomon proximity problems, which form the bedrock of many contemporary ZK-based technologies, including zk-rollups and zkVMs, and are critical to Ethereum’s long-term security roadmap, especially in a post-quantum era.
The better.codes platform takes a formalized problem from the Proximity Prize research, encapsulated in the Lean proof assistant, and places its current soundness bound onto a public leaderboard. Participants are invited to deploy their own AI agents and custom solvers to systematically increase this machine-checked soundness bound for "koalaIRS12," a specific Reed-Solomon proximity problem. The ultimate goal is to reach a target of 128 bits of security, a standard for many deployed SNARKs, by incrementally proving stronger security guarantees. Every successful submission, verified by the Lean kernel, advances the collective progress by raising the bound and contributing new lemmas, proof techniques, and potential impossibility results to a shared, publicly accessible repository.
The Crucial Role of Provable Security in Modern Cryptography
The genesis of better.codes lies in a critical need within the blockchain and cryptography communities: bridging the gap between conjectured security levels and rigorously proven security. Many of the advanced cryptographic primitives powering today’s decentralized systems, from the efficiency of zk-rollups to the future-proofing against quantum computers, rely heavily on Reed-Solomon codes and their associated proximity gaps and correlated agreement properties. These mathematical structures enable SNARKs to compress verifiable computations into small proofs, but their security guarantees are often based on complex mathematical conjectures.
"Nearly all production hash-based SNARKs, from the proof systems securing zkrollups and zkVMs to those central to Ethereum’s post-quantum roadmap, rely on proximity gaps and correlated agreement for Reed-Solomon codes," explains a spokesperson for the Ethereum Foundation’s Formal Verification team. "What can be proven about these results today stops short of what researchers believe the benchmarks may be. Deployed systems target 128-bit security, and that guarantee holds in full only if the conjectures do. The better.codes autoresearch challenge aims to close the gap between the conjectured security benchmarks and proven security benchmarks through open, incremental, verifiable, and public research."
This effort directly builds upon the Ethereum Foundation’s broader Proximity Prize initiative, launched earlier this year. The Proximity Prize aims to either definitively prove or disprove key conjectures surrounding Reed-Solomon proximity gaps. These grand challenges were notably outlined in the seminal paper "Open Problems in List Decoding and Correlated Agreement" by Gal Arnon, Dan Boneh, and Giacomo Fenzi. The specific problem addressed by better.codes, koalaIRS12, is derived from this foundational research and is fully formalized within ArkLib, the Lean 4 library for formally verified arguments of knowledge, ensuring a robust and precise mathematical foundation for all challenges.
An "Always-On" Autoresearch Paradigm
better.codes distinguishes itself through its "autoresearch" model, a novel approach to open-ended scientific discovery and collaborative problem-solving. Instead of a traditional competition with a single winner, better.codes functions as a continuous, parallelized research effort. Participants are encouraged to develop and deploy their own sophisticated AI agents, computational harnesses, and specialized tools. These independent systems then tackle the same verified benchmark problem concurrently.
"No single agentic setup is optimal across an open problem, so many independent setups working the same benchmark move the frontier faster than any one team can," stated a representative from Yukon, a key collaborator on the project. This distributed and parallel approach aims to accelerate progress by leveraging diverse strategies and computational resources. Previous challenges built on similar autoresearch principles, such as ecdsa.fail, zk.golf, and snark.fast, have already demonstrated significant success in advancing research frontiers in areas like quantum circuit design, verified ZK circuits, and optimizing SNARK proving speeds.
The philosophy behind this model is that complex, unsolved problems often benefit more from a multitude of diverse approaches than from a singular, centralized effort. By pooling the insights and discoveries from numerous independent solvers, the overall pace of research is amplified. Each successful proof, no matter how small an increment it represents, contributes to a more comprehensive understanding of the problem and builds upon the collective knowledge base.
How the better.codes Challenge Operates
The mechanics of participating in the better.codes challenge are designed for accessibility and transparency. Interested individuals or teams can sign up using their GitHub credentials on the better.codes website. Upon registration, they can clone the challenge repository, which contains the theorem statement, the initial parameter point, and the verification harness. Solvers then operate within a dedicated submission environment, aiming to prove a soundness lower bound that exceeds the current benchmark. The performance is measured in "bits of security," with each successful proof contributing to the overall progress.
A rigorous verification process is at the core of the challenge. A comparator ensures that each submitted proof’s exported theorem precisely matches the pinned statement. Crucially, the Lean kernel, a highly trusted formal verification engine, meticulously checks the correctness of every submitted proof. Accepted proofs are publicly recorded in the challenge’s repository, with full attribution given to the solver and the AI model or tools they employed.
This system of transparency and incremental contribution is fundamental to the autoresearch model. All submissions are git-backed, meaning every change, every lemma discovered, and every new proof technique is auditable and traceable. This allows participants to examine past diffs, study submission notes, learn from others’ successes and failures, and avoid redundant research paths. This continuous feedback loop fosters a collaborative environment where knowledge is shared openly, accelerating the collective journey towards the 128-bit security target.
Implications for the Future of Cryptography and Web3
The launch of better.codes represents a significant step forward in formal verification and the development of secure cryptographic primitives. By systematically improving the provable security of SNARKs, this challenge directly impacts the trustworthiness and robustness of numerous Web3 applications. zk-rollups, which are crucial for scaling Ethereum’s transaction throughput, rely on SNARKs for their security and efficiency. Strengthening the underlying cryptographic guarantees enhances the security of these scaling solutions, making them more resilient to potential attacks.
Furthermore, the focus on post-quantum security is prescient. As quantum computing capabilities advance, current cryptographic standards may become vulnerable. SNARKs based on Reed-Solomon codes are considered strong candidates for post-quantum cryptography. By ensuring their provable security at high bit levels now, the community is proactively building a foundation for a quantum-resistant future.
The autoresearch model itself offers a blueprint for tackling other complex, open-ended problems in computer science and mathematics. Its emphasis on distributed, verifiable, and transparent collaboration has the potential to democratize research and accelerate innovation across various fields. The success of better.codes could inspire similar initiatives in areas ranging from formal methods for AI safety to the development of new algorithmic designs.
What Lies Ahead
The immediate objective of the better.codes challenge is to elevate the proven lower bound for koalaIRS12 to the 128-bit security threshold. This is a substantial undertaking that will require significant ingenuity and computational effort from the participating solvers and their AI agents. The Ethereum Foundation and its collaborators have indicated that the platform is designed for expansion, with the potential to introduce additional challenges in the future, addressing other critical areas of cryptographic research.
Eligibility criteria, evaluation methodologies, and award structures are all governed by the program’s terms, which may be subject to adjustment as the challenge progresses and the community engages with it. The overarching goal remains to foster an environment of continuous, verifiable, and open research that benefits the entire ecosystem.
For researchers, developers, and anyone interested in the forefront of cryptographic security and decentralized technologies, better.codes offers a unique opportunity to contribute to a foundational element of the internet’s future. The platform is now live, inviting solvers to begin pushing the boundaries of provable security at https://better.codes.
