In an era where the security of global digital infrastructure relies increasingly on sophisticated cryptographic proofs, a new frontier has emerged in the intersection of artificial intelligence and formal verification. The Ethereum Foundation’s Formal Verification team, in a strategic partnership with Yukon and zkSecurity, has officially launched better.codes—an open, "always-on" autoresearch challenge designed to stress-test the foundations of modern zero-knowledge (ZK) systems.
By leveraging the rigorous, machine-checked language of Lean 4, the platform invites researchers and AI agents to collaborate on a high-stakes mission: bridging the chasm between theoretical security conjectures and provable mathematical guarantees. At its core, the platform challenges the community to raise the "soundness bound" of koalaIRS12—a critical Reed-Solomon proximity problem—toward a 128-bit target.
The Core Objective: Why Provable Bits Matter
The backbone of modern decentralized finance, including zkRollups, zkVMs, and Ethereum’s post-quantum roadmap, rests upon the integrity of hash-based SNARKs (Succinct Non-Interactive Arguments of Knowledge). These systems are frequently built upon conjectures involving proximity gaps and correlated agreement for Reed-Solomon codes.
However, a significant discrepancy currently exists in the field: the "proven" security of these systems often lags behind the "conjectured" security. While developers routinely deploy systems targeting 128-bit security—a standard benchmark for cryptographic safety—that guarantee is only as robust as the underlying mathematical proofs. If the conjectures governing these codes are not fully verified, the entire security model of the system remains vulnerable to theoretical edge cases.
The better.codes initiative is the first of its kind to treat these research gaps as a public, incremental, and machine-verified engineering problem. By shifting from static academic papers to a living, git-backed repository of verified proofs, the project aims to turn "conjectured" security into "demonstrable" security.
Chronology: From the Proximity Prize to the Lean Kernel
The origins of better.codes can be traced back to the Ethereum Foundation’s broader Proximity Prize initiative, launched earlier this year. The initiative was established to address the urgent need for a rigorous proof or refutation of Reed-Solomon proximity gap conjectures.
Key milestones in the development of this effort include:
- Early 2024: The launch of the Proximity Prize, focusing on incentivizing research into the most difficult open problems in list decoding and correlated agreement.
- The Academic Foundation: The publication of Open Problems in List Decoding and Correlated Agreement by researchers Gal Arnon, Dan Boneh, and Giacomo Fenzi, which provided the theoretical roadmap for the current challenge.
- Formalization: The integration of the problem into ArkLib, the Lean 4 library for formally verified arguments of knowledge, ensuring that every submission is subjected to the uncompromising scrutiny of the Lean kernel.
- The Present: The official go-live of the
better.codesinterface, allowing global contributors to submit proofs that are automatically checked, validated, and merged into the public record.
The Autoresearch Paradigm: A Collaborative Engine
The better.codes platform operates on the "autoresearch" model—a decentralized, agent-driven approach to solving complex scientific problems. Unlike traditional research, which is often siloed within individual universities or private laboratories, this platform encourages the parallel deployment of various AI agents and human-led research teams.
Why the "Always-On" Model Works
No single agentic configuration is currently optimized to solve all aspects of a complex cryptographic problem. By providing a common benchmark, the platform enables:
- Parallel Innovation: Different teams can test unique heuristics, LLM-based proof-generation strategies, and algorithmic search tools against the same pinned theorem.
- Cumulative Progress: When a participant submits a proof that raises the soundness bound, their new lemmas, proof techniques, and even impossibility results are "upstreamed." This means the next researcher does not have to start from scratch; they can build upon the most recent, successful breakthroughs.
- Transparency and Efficiency: By using a git-backed workflow, the platform acts as a public audit trail. Researchers can review past submissions, identify "dead ends" that others have already explored, and focus their computational resources on untapped areas.
This model draws inspiration from successful predecessors like ecdsa.fail, zk.golf, and snark.fast, which have already proven that public, competitive environments can accelerate progress in quantum circuit design and ZK-circuit optimization by years compared to traditional, closed-door research.
Supporting Data: The Mechanics of the Challenge
The challenge is centered on the koalaIRS12 parameter point. The workflow is designed for maximum security and reproducibility:
- GitHub Integration: Participants authenticate via GitHub to clone the challenge repository.
- The Constraint: The theorem statement and the verification harness are fixed. Participants must operate within a designated submission surface.
- The Metric: Submissions are scored in "bits." The primary goal is to iteratively increase the soundness lower bound of the system until it reaches the 128-bit threshold.
- Verification: A specialized comparator validates that the exported theorem matches the pinned challenge statement, and the Lean kernel performs the heavy lifting, verifying that the proof is logically sound.
This process ensures that "progress" is not merely an assertion but a mathematically guaranteed fact. Because Lean is a formal proof assistant, the platform effectively eliminates the possibility of "human error" or "hand-waving" in the proof process.
Official Perspectives: Implications for the Ecosystem
The launch of better.codes represents a paradigm shift for both the Ethereum Foundation and the broader cryptography community.
The Developer Perspective
For developers of ZK-based infrastructure, the platform provides a definitive answer to the question of "how secure is this really?" By formalizing these gaps, the industry can move away from relying on assumptions that are essentially "black boxes" and toward a future where every layer of the cryptographic stack is as verifiable as the software that implements it.
The Research Perspective
Academic researchers are provided with a novel way to gain recognition. By contributing lemmas and proof techniques that are integrated into the codebase, researchers are credited directly for their specific contributions. This aligns with a more open-science approach, where the "credit" for a breakthrough is shared by the collective effort of the community and the specific agent that unlocked the final piece of the puzzle.
Future Implications: Beyond the 128-Bit Target
While the immediate focus is on the 128-bit soundness bound for koalaIRS12, the creators of better.codes have signaled that this is merely the first chapter.
- Scaling Challenges: As the community succeeds in reaching the current goal, the platform is expected to host increasingly complex challenges, potentially covering other aspects of post-quantum cryptography, advanced polynomial commitments, and recursive SNARKs.
- Economic Incentives: The platform notes that eligibility, evaluation, and potential rewards are subject to evolving program terms. This suggests a future where researchers might be compensated for significant, machine-verified contributions to the cryptographic commons, creating a sustainable economy for open-source research.
- AI Integration: As AI agents become more adept at formal reasoning, the platform will likely become the primary training ground for "math-aware" AI. By providing a feedback loop where an agent can see its proof accepted or rejected by the Lean kernel, developers can fine-tune agents to become world-class formal verification engineers.
A New Standard for Digital Trust
The launch of better.codes is not just about a single math problem; it is about establishing a new standard for how we build digital trust. In a world where we increasingly rely on code to protect our most sensitive assets, the ability to "formally prove" the security of that code is no longer a luxury—it is a necessity.
By inviting the world to participate in this open, collaborative, and machine-verified research process, the Ethereum Foundation and its partners are ensuring that the future of cryptography is not built on shifting sands, but on the bedrock of formal, irrefutable mathematics.
For those looking to participate, the challenge is live now. The repository is open, the tools are ready, and the path to a more secure digital future is waiting to be written—one line of verified code at a time.
Visit better.codes to begin your contribution.
