At its core, better.codes is designed to tackle the inherent vulnerabilities in Succinct Non-Interactive Proofs of Knowledge, or SNARKs. These proof systems are the backbone of modern scaling solutions, including zero-knowledge rollups (zkRollups) and zero-knowledge virtual machines (zkVMs). As these technologies move from experimental research to the bedrock of financial and identity infrastructure, the margin for error effectively vanishes. The project invites researchers and AI-driven agents to collaborate on a public leaderboard, pushing the soundness bounds of the koalaIRS12 problem, a specific Reed-Solomon proximity challenge.
The Cryptographic Foundation of Modern Trust
To understand the significance of better.codes, one must first understand the reliance of modern cryptography on "proximity gaps." Most hash-based SNARKs currently in production rely on the assumption that certain Reed-Solomon codes maintain a level of correlated agreement. When developers build a zkRollup, they are essentially betting that the underlying mathematical conjectures regarding these codes hold true.
However, a discrepancy exists between what is currently proven and what is widely used. Industry standards for security generally target a 128-bit security level, a benchmark deemed sufficient to resist brute-force attacks by even the most powerful classical or near-future quantum computing architectures. Yet, in many cases, the rigorous, machine-checked proof for these 128-bit bounds is either incomplete or entirely absent. This creates a "security gap"—a zone where systems function correctly in practice but lack the ironclad formal verification that computer scientists demand.
The Ethereum Foundation, recognizing this, initiated the Proximity Prize earlier this year. The prize was established to incentivize the formal proof or refutation of these conjectures, as outlined in the landmark paper Open Problems in List Decoding and Correlated Agreement by researchers Gal Arnon, Dan Boneh, and Giacomo Fenzi. The better.codes challenge serves as the operational arm of this initiative, translating these high-level theoretical problems into actionable, code-based challenges that can be tackled incrementally.
The Autoresearch Model: A New Paradigm
The "autoresearch" model championed by better.codes is a departure from traditional academic silos. In conventional research, a team might spend months developing a proof, only for it to be peer-reviewed and published long after the initial discovery. In contrast, the autoresearch model treats the verification process as an open, continuous, and competitive pipeline.
By utilizing Lean 4—a robust, high-level functional programming language that serves as an interactive theorem prover—the project ensures that every submission is objectively verifiable. When a solver pushes a new lemma or proof technique, the Lean kernel automatically checks it for correctness. If the proof is valid and raises the soundness bound of the target problem, it is accepted to the leaderboard.
This creates a cumulative effect. Unlike traditional challenges where only the final winner receives recognition, better.codes archives every submission. New participants can review the history of attempts, study the "dead ends" that others have encountered, and build upon the lemmas already proven. This is, in effect, the democratization of mathematical proof-finding, accelerated by the integration of AI agents.
Chronology and Operational Flow
The launch of better.codes follows a trajectory of increasing interest in formal verification within the Ethereum ecosystem. The timeline of this movement can be traced back through several key milestones:
- Early 2024: The Ethereum Foundation intensifies its focus on post-quantum security and formal verification, leading to the creation of the Proximity Prize initiative.
- Mid-2024: The academic community releases updated benchmarks for list decoding and correlated agreement, highlighting specific areas where current proofs are insufficient.
- Late 2024: The Formal Verification team, in partnership with Yukon and zkSecurity, begins internal testing of the ArkLib library—a specialized Lean 4 library designed for formal arguments of knowledge.
- Present Day: The better.codes platform goes live, providing a public-facing interface for the koalaIRS12 problem.
The mechanics of participation are straightforward but demanding. Participants sign in via GitHub and clone the designated challenge repository. The environment is rigid; the theorem statement and the verification harness are fixed, ensuring that all submissions are evaluated against the exact same criteria. The leaderboard tracks progress in "bits," with the ultimate goal of reaching the 128-bit target.
Implications for the Future of ZK Infrastructure
The broader impact of this initiative extends far beyond the immediate resolution of the koalaIRS12 problem. If successful, the autoresearch model could become the standard for addressing "long-tail" problems in computer science—those that are too complex for a single researcher to solve but too modular to be ignored.
Furthermore, the integration of AI agents into the proof-finding process is a notable development. For years, the application of AI in mathematics was limited to heuristic pattern recognition. With better.codes, the role of AI shifts to the generation of verifiable, logical chains. If an AI can successfully navigate the state space of a complex mathematical problem and produce a result that the Lean kernel deems correct, it proves that the synergy between machine learning and formal logic is ready for large-scale application.
For the Ethereum ecosystem, the implications are even more direct. As the network looks toward its "post-quantum roadmap," the ability to formally verify the security of its proof systems is non-negotiable. If a vulnerability is found in a deployed SNARK, the impact could be catastrophic. By systematically raising the proven soundness of these systems, better.codes is performing a form of preventative security, reinforcing the digital infrastructure before the threat of quantum or advanced classical computation becomes a reality.
Expert Perspectives and Industry Reception
While the project is in its early stages, the reception from the cryptography community has been characterized by cautious optimism. Analysts note that while formal verification is notoriously difficult and time-consuming, the gamification of the process through leaderboards and prize incentives could lower the barrier to entry for highly skilled, yet unconventional, researchers.
"We are moving from a world where we hope our systems are secure based on peer-reviewed assumptions to a world where we know they are secure based on verified machine code," remarked one industry observer familiar with the project. "The shift is fundamental. It changes the nature of the software supply chain from one based on human trust to one based on mathematical certainty."
The collaboration between the Ethereum Foundation and specialized entities like Yukon and zkSecurity also underscores a growing trend of "mission-aligned" partnerships. By pooling resources and expertise, these organizations are creating public goods that benefit the entire ecosystem, rather than seeking to monopolize the underlying research.
What Comes Next: Scaling the Effort
As the better.codes platform gains traction, the organizers intend to expand the scope of the challenges. The current focus on koalaIRS12 is merely the first in what is expected to be a series of "grand challenges." As the proven security floor rises, the difficulty of finding the next marginal gain will increase, likely necessitating more sophisticated AI models and more robust collaborative efforts.
The program terms, which govern eligibility, evaluation, and potential rewards, are intentionally flexible. This allows the organizers to adjust the incentive structure as the landscape of the research evolves. For now, the call to action is clear: the repository is open, the problem is defined, and the leaderboard is waiting. Whether the 128-bit target is reached in months or years remains to be seen, but the process of reaching it will, in itself, provide invaluable insights into the limits and capabilities of modern formal verification.
In summary, better.codes represents a vital intersection of artificial intelligence, formal verification, and decentralized research. By codifying the search for mathematical truth, it moves the industry toward a more resilient, transparent, and provably secure future. For those interested in the frontier of cryptographic research, the platform offers a unique opportunity to contribute to a project that is, quite literally, defining the mathematical limits of the decentralized web.
