The Ethereum Foundation Formal Verification team, in partnership with research organizations Yukon and zkSecurity, has officially launched better.codes, a groundbreaking autoresearch challenge designed to crowdsource the formal verification of cryptographic soundness bounds. By utilizing the Lean theorem prover to host a public leaderboard, the initiative seeks to solve complex mathematical problems essential to the security of modern succinct non-interactive proof systems, or SNARKs.
At the core of the launch is the koalaIRS12 problem, a specific challenge related to Reed-Solomon proximity testing. This is not merely an academic exercise; it represents a significant effort to bridge the gap between theoretical security conjectures and machine-checked proofs that ensure the safety of infrastructure underpinning decentralized finance, zkRollups, and the broader Ethereum post-quantum roadmap.
The Cryptographic Foundation of Proximity Gaps
Modern cryptography relies heavily on the efficiency and security of hash-based SNARKs. These systems facilitate the verification of large-scale computations without requiring the verifier to re-execute the entire process. A critical component of these systems is the use of Reed-Solomon codes, which allow for efficient error correction and proximity testing—essentially ensuring that a given set of data is "close enough" to a valid codeword.
However, a fundamental tension exists between the practical performance of these systems and their theoretical security guarantees. Many deployed protocols operate under the assumption of 128-bit security, yet the underlying mathematical proofs for these bounds often rely on unproven conjectures. If these conjectures are found to be flawed, the security of the entire ecosystem could be compromised. The Proximity Prize initiative, launched by the Ethereum Foundation earlier this year, was specifically established to address these "proximity gaps."
The better.codes platform serves as the tactical implementation of this research strategy. By formalizing the koalaIRS12 problem within ArkLib—a specialized Lean 4 library—the organizers have provided a standardized, immutable environment where researchers and AI agents can compete to raise the proven security bound toward the 128-bit target.
A Chronology of Formal Verification Efforts
The movement toward formal verification in blockchain security has accelerated significantly over the past 24 months. The following timeline illustrates the evolution of these efforts:
- Early 2023: Increased scrutiny of zero-knowledge proof circuits highlights the risks associated with manual, human-checked code.
- Late 2023: The Ethereum Foundation identifies formal verification as a core pillar of its long-term security strategy, leading to the development of ArkLib.
- Early 2024: The Proximity Prize initiative is established, inviting global researchers to tackle the theoretical underpinnings of Reed-Solomon codes. This includes the publication of "Open Problems in List Decoding and Correlated Agreement" by renowned cryptographers Gal Arnon, Dan Boneh, and Giacomo Fenzi.
- Mid 2024: Integration of AI-driven research tools into the formal verification pipeline begins to show promise, demonstrating that agents can successfully automate portions of the proof-writing process.
- Late 2024: The launch of better.codes formalizes the "autoresearch" model, providing a public, git-backed interface for solvers to collaborate on the koalaIRS12 benchmark.
The Mechanics of the Autoresearch Challenge
The platform operates on a model of incremental, verifiable progress. When a participant signs in via GitHub, they gain access to the repository containing the pinned theorem statement for koalaIRS12. The objective is to produce a proof that the Lean kernel can verify, which effectively raises the "floor" of the security bound.
The process is transparent and highly iterative. Once a proof is submitted and passes the kernel verification, it is promoted to the public leaderboard. Because the submissions are tracked via git, the history of every proof—including failed attempts, new lemmas, and novel proof techniques—is preserved. This allows subsequent participants to learn from past failures, effectively creating a collaborative intelligence network.
The platform is designed to accommodate a diverse array of solvers. Some participants may utilize manual, human-authored proofs, while others are encouraged to deploy custom-built AI agents. The organizers anticipate that this "always-on" nature will create a competitive, yet collaborative, environment where the speed of innovation outpaces what any single research laboratory could achieve in isolation.
Why Formal Verification Matters for Global Infrastructure
The implications of the better.codes initiative extend far beyond the immediate technical challenge of the koalaIRS12 problem. In the current cybersecurity landscape, the "correctness" of software is increasingly tied to the "correctness" of the underlying math. Traditional peer review, while essential, is often too slow and error-prone for the rapidly evolving world of decentralized protocols.
By moving toward a model of machine-checked proofs, the Ethereum Foundation and its partners are setting a new standard for open-source research. If a proof is verified by the Lean kernel, it is mathematically certain that the result follows from the axioms, removing the ambiguity associated with human interpretation. This is particularly vital for post-quantum cryptography, where the complexity of the math is so high that traditional verification methods are increasingly insufficient.
Similar projects, such as ecdsa.fail and zk.golf, have already demonstrated the power of incentivized, open-source challenges in solving specific technical bottlenecks. By applying this "gamified" rigor to formal verification, the community is building a repository of verified knowledge that will serve as the foundation for the next generation of cryptographic proofs.
Expert Perspectives and Theoretical Challenges
While the enthusiasm for this initiative is high, the mathematical hurdles remain formidable. The "Open Problems in List Decoding and Correlated Agreement" paper serves as the north star for the project. The authors argue that the current understanding of proximity gaps in Reed-Solomon codes is incomplete, and that achieving 128-bit security requires a rigorous, systematic approach to proving lower bounds.
Industry observers note that the collaboration between the Ethereum Foundation, Yukon, and zkSecurity is significant because it aligns the incentives of theoretical researchers with those of protocol engineers. In previous eras, these groups often worked in silos. Today, however, the direct link between an abstract proof in Lean and the security of a production-grade zkVM is clear to all stakeholders.
The challenge, however, lies in the scalability of proof-writing. While AI agents are becoming more capable of writing code, they still struggle with the creative, heuristic leaps required to construct complex mathematical proofs. The better.codes platform acknowledges this by encouraging hybrid workflows, where human intuition guides the agent toward fruitful paths, while the agent handles the tedious, repetitive verification tasks.
Future Implications for the Ecosystem
The launch of better.codes marks a shift in how research is conducted in the blockchain space. By treating "research as code," the organizers are ensuring that progress is never lost. Each successfully verified proof adds a new layer of security to the protocols that rely on these mathematical foundations.
As the leaderboard matures, the project plans to introduce additional challenges. These will likely focus on other aspects of the cryptographic stack, including advanced zero-knowledge techniques, secure multi-party computation, and formal models of network security.
For developers and cryptographers, the message is clear: the future of secure software development lies in the marriage of advanced mathematics and automated, machine-checked verification. As participants begin to push the boundaries of the koalaIRS12 problem, they are not only competing for recognition and potential rewards—they are building the bedrock of a more secure and verifiable digital future.
For those interested in contributing to this effort, the challenge is now live at better.codes, where participants can begin by cloning the repository and reviewing the existing proof techniques. The platform welcomes contributions from across the global research community, regardless of their background in formal verification, provided they have the computational resources and mathematical interest to engage with the problem set.
The success of better.codes will ultimately be measured by how quickly it can close the gap between theoretical conjecture and proven security. In doing so, it may provide a template for how other complex, high-stakes engineering problems can be tackled through open, collaborative, and machine-verified research.
