The Ethereum Foundation’s Formal Verification team, in a strategic partnership with research firms Yukon and zkSecurity, has officially launched better.codes, an open-access "autoresearch" platform designed to accelerate the formal verification of cryptographic proof systems. By gamifying the advancement of machine-checked soundness bounds for Succinct Non-Interactive Arguments of Knowledge (SNARKs), the project seeks to bridge the gap between theoretical conjecture and verifiable mathematical certainty in the rapidly evolving landscape of zero-knowledge proofs.
The Cryptographic Security Imperative
At the heart of the initiative is the "koalaIRS12" challenge, which targets a critical component of modern blockchain infrastructure: Reed-Solomon proximity testing. Many of the most advanced proof systems currently securing zkRollups, zero-knowledge virtual machines (zkVMs), and post-quantum cryptographic roadmaps rely on specific assumptions regarding proximity gaps and correlated agreement.
While these systems are widely deployed, their security is often predicated on conjectures that remain unproven or are only partially understood. Industry standards typically target a 128-bit security level, a benchmark that ensures the mathematical impossibility of forging a proof within current computational limits. However, the theoretical foundations of these bounds are often narrower than the practical implementations suggest. By utilizing the Lean theorem prover—a rigorous language and environment for mathematical proofs—better.codes provides a mechanism to convert these loose conjectures into concrete, machine-verified theorems.
Chronology of the Proximity Prize Initiative
The launch of better.codes is the latest milestone in a broader, year-long effort by the Ethereum Foundation to fortify the mathematical underpinnings of the ecosystem. The timeline of this research trajectory reflects a growing emphasis on formal rigor:
- Early 2024: The Ethereum Foundation formally announces the Proximity Prize initiative, a research-heavy program dedicated to solving long-standing open problems in list decoding and correlated agreement.
- Mid-2024: Researchers Gal Arnon, Dan Boneh, and Giacomo Fenzi publish "Open Problems in List Decoding and Correlated Agreement," a foundational document detailing the grand challenges that must be addressed to secure next-generation SNARKs.
- Late 2024: ArkLib, a Lean 4-based library for verified arguments of knowledge, is integrated into the research workflow, providing a standardized environment for formalizing proof systems.
- Present: The activation of better.codes allows for a decentralized, crowdsourced approach to solving the specific koalaIRS12 problem, moving the needle on the 128-bit security target through incremental, iterative submissions.
The Autoresearch Model: A New Paradigm
The platform’s "autoresearch" model represents a departure from traditional, siloed academic research. Instead of relying on a single research team to solve a complex proof, better.codes invites independent participants to point their own artificial intelligence agents, specialized heuristics, and automated reasoning tools at the challenge.
This approach acknowledges that no single computational strategy is optimal for every mathematical problem. By creating a public, version-controlled leaderboard where every successful submission increases the "floor" of proven security, the project effectively turns global research efforts into a collective pipeline. When a solver submits a proof that passes the Lean kernel’s rigorous verification process, the new lemmas and proof techniques are upstreamed into the public repository. This ensures that future solvers do not repeat past mistakes, allowing the community to bypass dead ends and focus on progressively deeper layers of the mathematical structure.
Technical Context: Why Lean Matters
The reliance on Lean is not incidental. In the field of cryptography, a "proof" found in a PDF file or a peer-reviewed paper may still contain subtle, high-level logical errors that are difficult for human reviewers to catch. Formal verification eliminates this margin of error. When a proof is written in Lean, it is mathematically verified by a computer kernel. If the code compiles, the theorem is objectively true within the axioms defined.
For high-stakes infrastructure like Ethereum, this transition from "peer-reviewed" to "machine-verified" is a significant shift in risk management. The koalaIRS12 challenge requires solvers to navigate the complexities of Reed-Solomon proximity, specifically aiming to improve the soundness bound toward the 128-bit target. As participants iterate on these bounds, they are not merely performing a coding exercise; they are hardening the core security assumptions that underpin the future of private, scalable decentralized finance.
Industry Impact and Broader Implications
The implications of better.codes extend beyond the Ethereum ecosystem. As organizations and governments globally transition toward post-quantum cryptography, the demand for verified, high-assurance proof systems will only increase. Projects like ecdsa.fail, zk.golf, and snark.fast have previously demonstrated that open, incentive-aligned challenges can significantly accelerate innovation in circuit design and proving speed.
By integrating AI agents into this process, the Ethereum Foundation is tapping into a nascent trend of "AI-assisted mathematics." While AI models have historically struggled with the rigorous, error-intolerant nature of formal proof languages, tools like Lean are increasingly becoming the benchmark for evaluating AI’s logical reasoning capabilities. If better.codes succeeds in raising the soundness bound for koalaIRS12, it will provide a proof-of-concept for how human-AI collaboration can solve intractable problems in theoretical computer science.
Official Stance and Participation
The platform maintains a transparent, git-backed structure to ensure accountability. Participation is open to any researcher or engineer capable of interacting with the Lean environment. Each submission is credited to both the solver and the specific AI model or methodology used, fostering a culture of attribution and peer-to-peer knowledge sharing.
According to the organizers, the current focus on the 128-bit target is merely the beginning. The roadmap for better.codes suggests that once the current challenge matures, additional, more complex mathematical hurdles will be introduced. While the program’s terms regarding eligibility and potential rewards are managed by the Foundation, the primary incentive for many participants is the opportunity to contribute to a foundational, permanent artifact of cryptographic progress.
Conclusion
As the industry moves toward more complex and efficient proof systems, the need for formal verification is becoming paramount. better.codes addresses this need by democratizing the research process and applying the latest in automated theorem proving to some of the most pressing questions in cryptography. By turning the search for soundness into an open-source, collaborative challenge, the Ethereum Foundation and its partners are not just attempting to solve a single problem—they are building a new, more robust infrastructure for the future of digital verification. For those looking to participate, the repository is accessible via the better.codes portal, marking the beginning of an ongoing, public-facing effort to secure the mathematical foundations of the decentralized web.
