The Ethereum Foundation’s Formal Verification team, in a strategic partnership with Yukon and zkSecurity, has officially launched better.codes, an innovative "autoresearch" platform designed to bridge the gap between theoretical cryptographic conjectures and real-world security guarantees. This initiative introduces a public leaderboard where researchers and AI agents compete to solve formal verification problems, specifically targeting the soundness bounds of Reed–Solomon proximity—a foundational component of modern Succinct Non-Interactive Arguments of Knowledge (SNARKs).
The Genesis of the Challenge
The launch of better.codes follows the broader Proximity Prize initiative, which was established earlier this year to address the open questions surrounding Reed–Solomon proximity testing. At the heart of the current challenge is the "koalaIRS12" problem, a technical hurdle identified in the paper Open Problems in List Decoding and Correlated Agreement, authored by cryptographers Gal Arnon, Dan Boneh, and Giacomo Fenzi.
Reed–Solomon codes are the bedrock of many current zero-knowledge (ZK) proof systems, including those that power ZK-rollups and ZK-virtual machines. These systems often rely on the assumption that specific proximity gaps hold true under certain parameters. However, there is a persistent disparity between the security levels assumed in production environments—typically targeted at 128-bit security—and the security levels that have been mathematically proven. By formalizing these problems in the Lean 4 proof assistant, the better.codes platform seeks to systematically reduce this uncertainty.
The Mechanism of Autoresearch
The "autoresearch" model pioneered by better.codes represents a departure from traditional, siloed academic research. Instead of requiring a singular breakthrough from a lone researcher, the platform encourages a collaborative, iterative environment. Participants are invited to deploy their own AI agents, custom harnesses, and analytical tools to tackle the koalaIRS12 benchmark.
The workflow is designed for transparency and cumulative progress:
- Submission: Participants clone the challenge repository, which contains the pinned theorem statement and the Lean-based verification harness.
- Verification: Every submission is put through the Lean kernel, which acts as the ultimate arbiter. The kernel evaluates the proof, ensuring that the logic is sound and the soundness bound is indeed raised.
- Promotion: Successful proofs are promoted to the public repository. Importantly, these proofs are not merely "solved" and archived; they are upstreamed, meaning the new lemmas, proof techniques, and even documented "dead ends" become public resources for other participants.
This architecture ensures that no researcher starts from zero. As the collective pool of proofs grows, the difficulty of the next increment increases, effectively "ratcheting" the security floor of the entire ecosystem. This mirrors the methodology seen in previous successful collaborative research projects such as ecdsa.fail, zk.golf, and snark.fast, which have already significantly advanced the state of the art in quantum circuit design and post-quantum proving speeds.
Why Provable Bits Matter
In the world of blockchain infrastructure, "provable bits" are not just an academic abstraction; they are the primary metric of resistance against malicious actors. When a system is marketed as having "128-bit security," it implies that an adversary would need to perform approximately 2^128 operations to break the security guarantee.
However, many deployed systems rely on conjectures that have not been rigorously proven. If a conjecture is found to be weaker than expected, the actual security level of the protocol could be significantly lower than 128 bits. By using Lean, which provides machine-checked, computer-verified proofs, better.codes eliminates the ambiguity of human error in complex cryptographic proofs. The initiative is essentially a race to reach a mathematically verified 128-bit bound, ensuring that the cryptographic foundations of Ethereum’s post-quantum roadmap and its scaling solutions are as robust as they are theorized to be.
Chronology of Proximity Research
The path to better.codes can be traced back to the growing complexity of SNARKs and the increasing reliance on hash-based proof systems.
- Early 2024: The Ethereum Foundation formally announced the Proximity Prize initiative, signaling a commitment to funding rigorous, verifiable research into the mathematical foundations of ZK-proofs.
- Mid-2024: Publication of Open Problems in List Decoding and Correlated Agreement by Arnon, Boneh, and Fenzi, which codified the specific mathematical hurdles that currently impede full security verification.
- Late 2024: Development of ArkLib, the dedicated Lean 4 library designed for formally verified arguments of knowledge, which serves as the technical backbone for the better.codes platform.
- Present: The public release of the better.codes platform, allowing for decentralized, AI-augmented research into the koalaIRS12 problem.
Broader Implications and Industry Impact
The emergence of this platform suggests a paradigm shift in how high-stakes cryptographic research is conducted. By incentivizing open, agentic research, the Ethereum Foundation and its partners are effectively crowdsourcing the most difficult parts of formal verification. This approach is particularly effective for problems that are highly technical but discrete, where incremental progress can be easily verified and integrated.
For the wider cryptographic community, the implications are significant. First, the standardization of proofs in Lean creates a "common language" for security researchers. Second, the use of AI agents to explore the state space of proofs accelerates the identification of efficient, high-performance proof techniques. Finally, the project demonstrates a model of "always-on" research, where the progress of the field is not dictated by conference cycles or quarterly grant deadlines, but by the continuous, automated efforts of a global community.
While the current scope is limited to the koalaIRS12 problem, the infrastructure of better.codes is designed to be extensible. The platform’s organizers have indicated that future challenges will be added as the current one approaches its 128-bit target. This creates a long-term roadmap for formal verification, ensuring that as new cryptographic primitives are developed, their security bounds can be verified with the same level of rigor.
Official Stance and Future Governance
While the technical aspects of the challenge are decentralized in their execution, the program remains governed by specific terms regarding eligibility, evaluation, and rewards. Payments and awards are designed to incentivize high-quality, reproducible contributions, but they are subject to change as the challenge matures.
The collaboration between the Ethereum Foundation’s Formal Verification team, Yukon, and zkSecurity underscores a consensus that the "black box" nature of cryptographic security is no longer sustainable. As the ecosystem moves toward increasingly complex ZK-proof systems, the ability to mathematically verify these systems is a prerequisite for long-term survival. Through better.codes, the research community is attempting to move from a state of "security by conjecture" to "security by proof," a transition that will likely define the next decade of blockchain development.
Researchers, developers, and AI practitioners interested in contributing to this effort are encouraged to visit the better.codes website, where the full documentation, the ArkLib library, and the submission guidelines are currently available for public engagement. As the challenge proceeds, the platform will serve as a living archive of how the industry successfully navigated one of its most persistent and critical mathematical bottlenecks.
