In a bold move to bridge the gap between theoretical cryptography and real-world security, the Ethereum Foundation’s Formal Verification team has launched better.codes. Developed in close collaboration with the research organizations Yukon and zkSecurity, the platform introduces a novel "autoresearch" challenge designed to harden the foundations of modern zero-knowledge proof systems.
By leveraging the Lean theorem prover to host a public, crowd-sourced competition, the initiative aims to push the boundaries of what is mathematically provable in Reed–Solomon proximity testing. For the blockchain ecosystem, this is not merely an academic exercise; it is an effort to move beyond conjecture and solidify the 128-bit security guarantees that underpin the next generation of zk-Rollups, virtual machines, and post-quantum cryptographic roadmaps.
Main Facts: The Intersection of AI and Formal Verification
The better.codes platform functions as an open-access, machine-checked leaderboard for cryptographic research. At its core is the koalaIRS12 problem, a specific Reed–Solomon proximity challenge derived from the broader Proximity Prize initiative.
Participants are tasked with utilizing their own AI agents, custom harnesses, and computational tools to increase the "soundness bound" of the koalaIRS12 problem. In cryptography, a soundness bound defines the probability that an adversary can trick a proof system into accepting a false statement. Currently, many production-grade systems rely on conjectures that claim to offer 128-bit security; however, the mathematical proofs backing these claims often lag behind the deployed implementations.
The better.codes platform requires every submission to be validated by the Lean kernel. This ensures that no claim is accepted without rigorous, machine-verifiable proof. Once a submission is verified, it is added to the public leaderboard, and its underlying lemmas, proof techniques, and even identified impossibility results are upstreamed to a shared repository. This transparency ensures that the community is not just competing, but iteratively building a collective knowledge base.
Chronology: From Proximity Prize to Public Challenge
The journey toward better.codes began with the identification of a significant vulnerability in the collective understanding of cryptographic proofs. Throughout 2024, the Ethereum Foundation recognized that as zero-knowledge (ZK) infrastructure scales, the reliance on unproven conjectures could represent a systemic risk.
Phase 1: The Proximity Prize
Earlier this year, the Foundation launched the Proximity Prize initiative, aimed at solving fundamental open questions in list decoding and correlated agreement. This initiative was heavily influenced by the seminal paper, “Open Problems in List Decoding and Correlated Agreement,” co-authored by researchers Gal Arnon, Dan Boneh, and Giacomo Fenzi. The paper identified the exact "proximity gaps" that remain the "holy grail" of ZK research.
Phase 2: Formalizing the Problem
Following the publication of these open problems, the Formal Verification team worked to translate these complex mathematical hurdles into ArkLib, a specialized Lean 4 library designed for formally verified arguments of knowledge. By encoding the problem in Lean, the team provided a rigid, immutable target that could be tested by automated agents.
Phase 3: The Launch of better.codes
With the challenge environment stabilized, better.codes went live as a permanent, always-on infrastructure. The platform draws inspiration from previous successful research-acceleration projects like ecdsa.fail, zk.golf, and snark.fast. Each of these predecessors demonstrated that by standardizing the "rules of the game" and exposing the frontier of a problem, the global research community can arrive at solutions significantly faster than any isolated academic team.
Supporting Data: Why "Provable Bits" Matter
To understand the necessity of better.codes, one must look at the state of contemporary cryptographic security. Most production-grade, hash-based SNARKs—the systems that allow Ethereum to verify large amounts of data without processing every transaction individually—rely on specific mathematical properties of Reed–Solomon codes.
The Security Gap
- The Target: 128-bit security is the industry standard for a system that is computationally infeasible to break with current and near-future technology.
- The Reality: Many deployed systems operate on the assumption that these 128-bit guarantees hold. However, the formal proofs that verify these claims are often incomplete or based on "conjectured" behavior rather than absolute mathematical certainty.
- The Mechanism: The koalaIRS12 problem represents a specific parameter point. By incrementally raising the proven soundness bound through better.codes, the platform systematically eliminates the "conjecture" and replaces it with "proof."
The Multiplier Effect
The power of this initiative lies in its "agentic" approach. Because no single AI architecture or human team possesses the perfect heuristic for solving complex, high-dimensional proof problems, the platform encourages diversity. By running thousands of independent setups in parallel, the community creates a high-throughput pipeline of proofs. When one agent discovers a new lemma that raises the bound by even a fraction, that breakthrough becomes public knowledge, allowing every other agent to utilize that proof in their own next iteration.
Official Responses and Strategic Implications
The launch of better.codes has been met with optimism from both the formal methods community and the Ethereum developer ecosystem.
Bridging the "Trust" Chasm
In a statement regarding the launch, contributors from the Ethereum Foundation noted that the goal is not just to "win" a competition, but to raise the floor of security for the entire blockchain space. "We are moving from a world where security is a matter of expert consensus to a world where it is a matter of machine-verified fact," one lead researcher suggested.
The Role of Yukon and zkSecurity
The involvement of external firms like Yukon and zkSecurity highlights the collaborative nature of modern Ethereum research. By outsourcing the challenge infrastructure to experts in formal verification and ZK-security, the Ethereum Foundation ensures that the platform remains neutral and rigorous. These organizations are responsible for ensuring that the Lean proofs are not only accurate but also interoperable with other emerging cryptographic standards.
Global Research Implications
The implications for the broader field of computer science are profound. If the autoresearch model succeeds in closing the security gaps in SNARKs, the methodology will likely be exported to other domains:
- Post-Quantum Cryptography: Identifying which algorithms can survive quantum-enabled attacks.
- Verified Hardware Design: Using the same "always-on" approach to verify the security of physical circuit designs.
- Automated Theorem Proving: Setting a new standard for how researchers interact with LLMs and AI-driven proof assistants.
The Path Forward: How to Engage
The better.codes platform is designed to be accessible to any researcher or agent-based system capable of interacting with the GitHub-backed challenge repository.
Participation Mechanics
- Authentication: Users authenticate via GitHub, ensuring that all submissions are traceable and attributed.
- The Submission Loop: After cloning the repository, solvers work within a designated "submission surface." The process involves writing Lean code to increase the soundness lower bound of the koalaIRS12 problem.
- Validation: A comparator checks the exported theorem against the pinned problem statement. The Lean kernel then performs a deep check of the logic.
- The Ledger of Progress: Successful submissions are committed to the public ledger. This creates a permanent, immutable history of the research, where every "diff" tells the story of how the security bound was pushed forward.
Future Challenges
While the current focus is squarely on the 128-bit target for koalaIRS12, the architecture of better.codes is built for extensibility. The organizers have indicated that they intend to roll out further challenges as the community masters the current one. The program terms, including eligibility and potential rewards for breakthroughs, are designed to evolve alongside the research, ensuring that the platform remains incentivized and responsive to the needs of the ZK community.
Conclusion: A New Era for Mathematical Certainty
The launch of better.codes represents a transition in how we think about cryptographic security. We are moving away from the "black box" era—where we rely on the intuition of a few elite cryptographers—and toward a "glass box" era, where every line of code is held to the highest standard of formal verification.
By inviting the world to tackle these problems in an open, competitive, and collaborative environment, the Ethereum Foundation is effectively "crowdsourcing the truth." As AI agents and human researchers begin to feed their results into the leaderboard, the resulting progress will do more than just improve a single protocol; it will provide the foundational security upon which the next decade of decentralized computing will be built.
For those looking to contribute to the cutting edge of mathematics and cybersecurity, the challenge is now live at better.codes.
