In a significant leap for the intersection of artificial intelligence, formal verification, and cryptographic infrastructure, the Ethereum Foundation’s Formal Verification team—in collaboration with research partners Yukon and zkSecurity—has officially unveiled better.codes. This ambitious initiative represents a new paradigm in "autoresearch": a public, open-ended challenge designed to accelerate the mathematical proof of security benchmarks that underpin the next generation of decentralized technologies.
By utilizing the Lean theorem prover to formalize complex cryptographic problems, better.codes invites researchers, AI agents, and cryptographers to compete in a transparent, collaborative environment. The objective is to push the boundaries of what is mathematically provable in succinct non-interactive proof systems (SNARKs), effectively hardening the bedrock upon which modern zkRollups, virtual machines, and post-quantum cryptographic roadmaps are built.
Main Facts: The Intersection of Lean and Cryptography
At the heart of the better.codes platform lies the koalaIRS12 problem. This is not merely a theoretical exercise; it is a critical component of Reed-Solomon proximity research. Modern SNARKs, which are essential for scaling Ethereum and ensuring the privacy of user transactions, rely heavily on these codes.
However, there is a fundamental tension in the industry: while production systems often operate under the assumption of 128-bit security, the formal mathematical proofs supporting these assumptions often trail behind the practical deployment of the code. The better.codes platform seeks to bridge this gap. Participants are tasked with using AI-driven solvers to raise the "machine-checked soundness bound" of the koalaIRS12 problem.
Every submission is subjected to the rigor of the Lean kernel. Unlike traditional software development, where unit tests provide a baseline of functionality, Lean provides a mathematical guarantee of correctness. When a solver pushes a new proof, it is verified by the kernel and, if successful, permanently raises the floor of the global benchmark. This creates a cumulative knowledge base where every participant builds directly upon the discoveries of their predecessors.
Chronology: From Proximity Prize to Public Challenge
The launch of better.codes is the culmination of a broader strategic shift within the Ethereum Foundation toward more rigorous, open-source research methodologies.
The Proximity Prize Genesis
The trajectory began earlier this year with the launch of the Proximity Prize initiative. Recognizing that the security of many ZK-based systems relied on unproven conjectures, the Foundation sought to incentivize the academic and developer communities to address these open problems. The initiative was grounded in the seminal work of Gal Arnon, Dan Boneh, and Giacomo Fenzi, whose paper, “Open Problems in List Decoding and Correlated Agreement,” outlined the grand challenges facing the field.
Formalization in ArkLib
Following the identification of these challenges, the researchers moved to formalize the problems within ArkLib, a specialized Lean 4 library designed for verified arguments of knowledge. By encoding these problems in a language that computers can verify, the team ensured that any solution found would be beyond dispute.
The Launch of the Platform
Following months of development, the better.codes platform went live, integrating GitHub authentication to allow for a transparent, git-backed submission process. This transition from static paper-based research to a dynamic, "always-on" platform marks a departure from traditional academic cycles, favoring a continuous, collaborative, and competitive approach to solving complex problems.
Supporting Data: Why "Provable Bits" Matter
The necessity for better.codes is driven by the gap between empirical confidence and formal certainty. In cryptography, "security" is often measured in bits—a 128-bit security level implies that an attacker would need $2^128$ operations to break the system. While this is the industry gold standard, it is frequently based on conjectures rather than ironclad proofs.
The Reed-Solomon Gap
Reed-Solomon codes are the workhorses of ZK-SNARKs. They allow a prover to convince a verifier that they know a secret polynomial without revealing it. The "proximity gap" refers to the difficulty of proving that a received word is "close" to a Reed-Solomon code. If the bound on this proximity is weaker than expected, the system could be vulnerable to sophisticated attacks that the original designers did not anticipate.
The "Autoresearch" Model
The efficacy of the better.codes model is supported by the success of its predecessors:
- ecdsa.fail: Pushed the limits of understanding regarding Elliptic Curve Digital Signature Algorithm implementations.
- zk.golf: Focused on the optimization of ZK circuits, proving that independent agents can find efficiencies that human researchers often overlook.
- snark.fast: Advanced the state of the art in post-quantum proving speeds.
By crowdsourcing the intellectual labor of finding new lemmas, proof techniques, and impossibility results, better.codes effectively turns the research process into a massive parallel computing project where the "compute" is high-level mathematical reasoning.
Official Responses and Strategic Vision
The collaboration between the Ethereum Foundation, Yukon, and zkSecurity is indicative of a broader industry consensus: the future of blockchain security depends on the formalization of its mathematical foundations.
In statements regarding the platform, the organizers emphasized that no single team has a monopoly on the optimal approach to these proofs. "No single agentic setup is optimal across an open problem," the team noted in their launch documentation. By fostering a environment where independent agents compete on the same benchmark, the platform ensures that the "frontier" of knowledge moves faster than any single team—no matter how well-funded—could manage alone.
Furthermore, the platform emphasizes the importance of upstream contributions. When a participant finds a new way to prove a subset of the problem, that lemma is not kept as a competitive secret. Instead, it is integrated into the repository, allowing all other agents to skip "dead ends" and focus their computational efforts on higher-level problems. This creates a "rising tide" effect where the entire community benefits from individual successes.
Implications: The Future of Formal Verification
The launch of better.codes carries significant implications for both the blockchain industry and the broader field of computer science.
1. The Rise of Agentic Research
We are witnessing a shift where human researchers act as architects, while AI agents act as the primary laborers of mathematical proof. The ability for an AI to interact with a Lean kernel, receive feedback, and iterate on a proof in real-time is a preview of how scientific discovery will likely function in the coming decade.
2. Hardening Ethereum’s Infrastructure
As Ethereum continues its evolution toward a more modular, ZK-heavy architecture, the "provability" of its underlying protocols becomes a systemic risk factor. By systematically raising the soundness bounds of these systems, better.codes acts as a long-term insurance policy for the Ethereum ecosystem. The results generated here will directly inform the security parameters of future Layer 2 rollups and ZK-VMs.
3. Democratization of Cryptographic Research
By providing a web-based, git-backed interface, the platform lowers the barrier to entry for cryptographic research. One does not need to be a tenured professor or a lead researcher at a top-tier lab to contribute to the foundations of ZK-SNARKs. A high school student with a strong grasp of Lean and a clever AI model could, in theory, find the breakthrough that secures a billion-dollar protocol.
4. A New Standard for "Open"
The transparency of the better.codes platform serves as a blueprint for other high-stakes software sectors. In a world where digital infrastructure is often proprietary and opaque, the commitment to open, incremental, and verifiable progress offers a refreshing alternative. It proves that public goods can be built through competitive, high-velocity collaboration.
Moving Forward
As the challenge progresses, the Ethereum Foundation plans to introduce further problems, expanding the scope of the platform beyond Reed-Solomon proximity. While the current focus is on reaching that 128-bit threshold for koalaIRS12, the platform is designed to be a permanent fixture. Eligibility and awards are governed by program terms, ensuring that the incentive structure remains aligned with the technical goals.
For researchers, cryptographers, and AI enthusiasts, the call to action is clear: log in, clone the repository, and start proving. In the world of better.codes, the math is the only thing that matters, and the frontier is waiting to be pushed. The work done today is the foundation for the secure, scalable, and decentralized digital world of tomorrow.
