In a landmark move for the future of cryptographic security, the Ethereum Foundation’s Formal Verification team, in partnership with Yukon and zkSecurity, has officially launched better.codes. This platform serves as a high-stakes, open-source "autoresearch" challenge designed to bridge the critical gap between theoretical cryptographic conjectures and the rigorous, machine-checked security required for modern blockchain infrastructure. By leveraging the Lean theorem prover, the initiative invites researchers and AI agents to solve complex problems in Reed-Solomon proximity, fundamentally strengthening the backbone of zero-knowledge proof systems (SNARKs).
Main Facts: A New Paradigm for Cryptographic Research
At its core, better.codes is an experimental, collaborative environment where the traditional, siloed pace of academic research is replaced by a high-speed, iterative, and public-facing methodology. The platform revolves around the "koalaIRS12" problem—a specific challenge rooted in the Proximity Prize research initiative.
The primary objective is simple yet profoundly difficult: to raise the proven soundness bound of the koalaIRS12 Reed-Solomon proximity problem toward a target of 128 bits. In the world of cryptography, 128-bit security is the gold standard for resisting brute-force attacks. Currently, many production-grade systems rely on assumptions that have yet to be formally proven. better.codes changes this by requiring every submission to be verified by the Lean kernel, ensuring that every "bit" of security added to the leaderboard is mathematically undeniable.
Participants—whether human researchers or automated AI agents—clone the challenge repository, work within a defined submission surface, and push their verified proofs to a public leaderboard. Each successful submission does more than just earn a spot on the board; it contributes new lemmas, novel proof techniques, and documented "dead ends" to the repository, creating a cumulative knowledge base that accelerates progress for every subsequent solver.
Chronology: From Academic Conjecture to Public Challenge
The trajectory leading to the launch of better.codes began with the recognition that while SNARKs (Succinct Non-interactive Arguments of Knowledge) are becoming the primary mechanism for scaling Ethereum, their security foundations remained "conjectural."
Earlier this year, the Ethereum Foundation formalized its commitment to this issue by launching the Proximity Prize initiative. This project was specifically tasked with proving or disproving the Reed-Solomon proximity gap conjectures—the very math that determines whether a proof system can be subverted by malicious actors.
The release of the paper, Open Problems in List Decoding and Correlated Agreement by Gal Arnon, Dan Boneh, and Giacomo Fenzi, served as the intellectual blueprint. This paper highlighted the deep mathematical challenges inherent in list decoding, providing the "grand challenges" that better.codes is now operationalizing. By integrating these problems into ArkLib—a dedicated Lean 4 library for formally verified arguments of knowledge—the organizers moved the research from the abstract realm of PDFs and seminars into the practical, executable realm of code.
Following the initial development phase, the team integrated the "autoresearch" model, drawing inspiration from prior community successes like ecdsa.fail, zk.golf, and snark.fast. These precursors demonstrated that when you gamify complex mathematical problems and provide a public, transparent leaderboard, the community can solve problems in months that might otherwise take years of traditional academic peer review.
Supporting Data: Why "Provable Bits" Matter
To understand the necessity of better.codes, one must look at the current state of zero-knowledge proofs. Today’s zkRollups and zkVMs (Zero-Knowledge Virtual Machines) rely heavily on hash-based SNARKs. The integrity of these systems depends on the assumption that if an attacker provides a "near" codeword, they cannot easily deceive the verifier.
However, researchers have long known that the gap between what is theoretically conjectured and what is formally proven is a potential point of failure. If the proximity gap is smaller than anticipated, the security of the entire network could be at risk.
The "koalaIRS12" challenge represents a specific configuration of this problem. A 128-bit security target means that an attacker would need to perform $2^128$ operations to break the security, a computationally infeasible task with current technology. Currently, many systems assume this level of security, but the mathematical proof for the specific parameters used in these SNARKs is often incomplete. By pushing the proof toward 128 bits, better.codes is essentially hardening the "bedrock" upon which Ethereum’s post-quantum roadmap is built.
The leaderboard acts as a barometer for this security. Every time a solver improves the bound, the "proven security" of the underlying math increases, moving closer to the "conjectured security" that developers have been relying on.
Official Responses and Collaborative Philosophy
The design philosophy behind better.codes is one of "Agentic Collaboration." The organizers emphasize that no single human researcher or AI model is likely to hold the optimal strategy for such a difficult mathematical problem. By creating an "always-on" environment, the organizers encourage a diversity of approaches.
"The goal is to move the frontier faster than any one team could," said a spokesperson for the initiative. "By upstreaming every new lemma and proof technique, we ensure that the entire community benefits from the breakthroughs of the few. It’s not just about winning the prize; it’s about collectively exhausting the solution space."
The partnership between the Ethereum Foundation, Yukon, and zkSecurity reflects a strategic alliance between protocol stewards, formal verification specialists, and security researchers. By opening the challenge to anyone with a GitHub account, the organizers are democratizing a field that has historically been limited to a small group of high-level cryptographers.
Implications: The Future of Formal Verification
The launch of better.codes carries significant implications for the broader blockchain ecosystem and beyond:
1. The Rise of "Agentic" Research
We are witnessing a shift where human researchers act as architects, designing the "harnesses" for AI agents to conduct the heavy lifting of formal verification. As these AI models become more adept at interacting with Lean, the speed of mathematical discovery is expected to increase exponentially. This is a bellwether for how science, particularly complex mathematics, will be conducted in the future.
2. Standardizing Security
By creating a common, public benchmark, better.codes is creating a standard. If a SNARK implementation cannot meet the security bounds established on the better.codes leaderboard, it will become increasingly difficult for developers to claim their systems are "secure" in a professional or production capacity.
3. Strengthening the Ethereum Roadmap
The Ethereum roadmap is heavily reliant on cryptography that is performant, scalable, and—most importantly—correct. The better.codes platform provides a vital feedback loop. If the research reveals that certain parameters are inherently insecure, it allows protocol developers to pivot before those vulnerabilities are deployed in live, high-value networks.
4. Beyond Blockchain
While the immediate beneficiaries are the users of Ethereum’s zk-based scaling solutions, the results from better.codes have implications for any field that requires high-assurance software. The techniques developed for koalaIRS12 could eventually inform the verification of hardware, cryptographic protocols in traditional finance, and even AI safety, where formal verification is increasingly viewed as a necessary guardrail.
Conclusion: How to Participate
The challenge is currently live and open for participation. Interested solvers are invited to visit better.codes to access the challenge documentation, the ArkLib library, and the submission portal.
Participation is governed by transparent program terms, which account for the evolution of the research. As the community continues to chip away at the 128-bit target, the repository will grow, serving as a living history of the effort to make the digital world more secure. For those looking to be at the bleeding edge of formal verification, the better.codes challenge offers more than just a leaderboard—it offers the chance to participate in the formalization of the very math that will secure the next generation of the internet.
As the industry moves toward a future defined by zero-knowledge proofs, the work happening today at better.codes will likely be viewed as the moment the industry stopped "hoping" for security and started "proving" it.
