The Ethereum Foundation, in a significant stride towards bolstering the foundational security of cryptographic systems, has officially launched "better.codes," an innovative open autoresearch challenge. This initiative, a collaborative effort with Yukon and zkSecurity, aims to push the boundaries of verifiable security in modern proof systems, particularly those underpinning zero-knowledge technologies and future blockchain applications. The platform introduces a novel approach to research, leveraging AI agents and collective intelligence to incrementally improve cryptographic guarantees.
The Genesis of better.codes: A Commitment to Provable Security
The inception of better.codes is rooted in the ongoing pursuit of stronger, provable security guarantees within the realm of cryptography, especially concerning succinct non-interactive arguments of knowledge (SNARKs). SNARKs are crucial for technologies like zk-rollups and zkVMs, enabling efficient and private transaction processing on blockchains, and are also a cornerstone of Ethereum’s roadmap for post-quantum cryptography. A core component of many of these SNARKs involves Reed-Solomon proximity problems and correlated agreement. However, the theoretical security bounds of these systems often outpace what has been rigorously proven.
The Proximity Prize initiative, launched earlier this year by the Ethereum Foundation, directly addresses this gap. This initiative, detailed in the seminal paper "Open Problems in List Decoding and Correlated Agreement" by Gal Arnon, Dan Boneh, and Giacomo Fenzi, seeks to either prove or disprove key conjectures related to Reed-Solomon proximity gaps. better.codes serves as a practical, gamified extension of this research, providing a public arena for participants to contribute to raising the "soundness bound" of a specific, formalized problem.
The koalaIRS12 Challenge: A Formalized Frontier
At the heart of the better.codes platform is the "koalaIRS12" challenge. This challenge originates from a specific problem within the Proximity Prize research, meticulously formalized in Lean, a powerful proof assistant. Lean is renowned for its ability to formally verify mathematical proofs, ensuring an exceptionally high degree of rigor and correctness. The koalaIRS12 problem is a Reed-Solomon proximity problem, a type of mathematical puzzle that, when solved, yields crucial insights into the efficiency and security of SNARKs.
The objective for participants is to direct their AI agents, or any other suitable computational tools, towards enhancing the machine-checked soundness bound of koalaIRS12. This bound represents a measure of security, and the ultimate target is a robust 128-bit security level, a standard widely adopted in production SNARKs. Every successful submission that improves this bound is publicly recognized on a leaderboard, fostering a competitive yet collaborative research environment.
An "Always-On" Autoresearch Model: Collective Intelligence in Action
better.codes operates under an "always-on autoresearch" model, a paradigm shift in how research challenges are conducted. This model encourages a decentralized and parallel approach to problem-solving. Instead of relying on a single team or a centralized effort, better.codes invites participants to deploy their unique AI agents, harnesses, and custom tools. These independent efforts are then pitted against the common benchmark, allowing for a diverse range of approaches and accelerating the discovery of novel solutions.
This decentralized approach mirrors the success of previous open challenges orchestrated by similar entities, such as ecdsa.fail, zk.golf, and snark.fast. These initiatives have demonstrably pushed research frontiers in areas like quantum circuit design, verified ZK circuits, and post-quantum proving speeds. The underlying principle is that no single algorithmic approach is universally optimal for complex research problems. By enabling numerous independent agents to tackle the same verified benchmark simultaneously, the collective progress is significantly amplified.
The Mechanics of Contribution: Incremental, Verifiable, and Transparent
The process on better.codes is designed for clarity, verifiability, and incremental progress. Participants begin by signing in with their GitHub accounts, a standard practice for open-source collaboration, and cloning the challenge repository. This repository contains the core elements of the challenge: the theorem statement, the specific parameter point being investigated, and a verification harness that ensures the integrity of submissions.
Solvers then work within a dedicated submission environment. Their task is to prove a higher soundness lower bound for koalaIRS12, with the score measured in bits of security. Each submission undergoes a rigorous vetting process. A comparator first verifies that the exported theorem from the submission precisely matches the pinned statement. Subsequently, the Lean kernel, the engine of formal verification, meticulously checks the submitted proof.
Accepted submissions are not only added to the public repository but are also credited to the solver and the AI model used. This transparency is crucial. All submissions are git-backed, meaning every change, every lemma discovered, and every proof technique employed is recorded and auditable. This allows any participant to examine past diffs, read submission notes, learn from successful strategies, and crucially, avoid repeating unsuccessful paths. This continuous upstreaming of new lemmas, proof techniques, and even impossibility results ensures that all solvers and their agents benefit from the collective advancements, creating a virtuous cycle of research progress.
Why Provable Bits Matter: The Stakes for Modern Cryptography
The emphasis on "provable bits" is not merely an academic pursuit; it has profound implications for the security of widely deployed cryptographic systems. Nearly all production hash-based SNARKs, the very systems securing zk-rollups, zkVMs, and forming the bedrock of Ethereum’s long-term security strategy against quantum computing threats, rely on the mathematical underpinnings of proximity gaps and correlated agreement within Reed-Solomon codes.
The security offered by these deployed systems is typically quantified at 128 bits, a level considered robust against current and foreseeable computational capabilities. However, this guarantee is contingent upon the validity of certain mathematical conjectures. In essence, the full 128-bit security is only realized if these underlying assumptions hold true. The challenge of better.codes directly addresses this critical juncture: closing the gap between conjectured security benchmarks and rigorously proven security benchmarks. By fostering open, incremental, verifiable, and public research, the platform aims to elevate the certainty of these security claims.
The formalization of koalaIRS12 in ArkLib, the Lean 4 library dedicated to formally verified arguments of knowledge, bridges the abstract research challenges outlined in "Open Problems in List Decoding and Correlated Agreement" to a concrete, solvable problem. This ensures that progress made on better.codes has direct relevance and impact on the broader Proximity Prize initiative and, by extension, on the security of critical blockchain infrastructure.
Looking Ahead: Expanding the Frontier of Verified Cryptography
The initial launch of better.codes focuses on the soundness challenge for koalaIRS12, with the explicit goal of raising its proven lower bound to the target of 128 bits. However, the vision extends beyond this single challenge. The Ethereum Foundation anticipates adding further research problems and challenges to the platform over time, continually expanding the scope of verifiable cryptographic research.
The program’s terms, including eligibility criteria, evaluation methodologies, awards, and payment structures, are clearly defined and may be subject to adjustments as the challenge evolves. This adaptive approach ensures that the platform remains responsive to the dynamic nature of research and the collective progress of its participants.
The launch of better.codes represents a significant step forward in the quest for demonstrably secure cryptographic systems. By harnessing the power of open collaboration, AI, and formal verification, the Ethereum Foundation and its partners are cultivating an environment where research is not only accelerated but also made more transparent, verifiable, and ultimately, more trustworthy. The platform invites researchers, developers, and AI enthusiasts alike to contribute to this vital effort, helping to build a more secure digital future. Interested parties can begin their journey at better.codes.
