The landscape of zero-knowledge cryptography and blockchain scalability has taken a significant step forward with the official launch of better.codes. Developed as an open autoresearch challenge by the Ethereum Foundation Formal Verification team in close collaboration with Yukon and zkSecurity, the initiative invites researchers and developers to leverage autonomous artificial intelligence agents to tackle complex mathematical problems foundational to modern succinct non-interactive proof systems, commonly known as SNARKs.
At its core, better.codes introduces a self-contained problem derived from the Proximity Prize research initiative. Formalized entirely within the Lean interactive theorem prover, the challenge places its soundness bound on a public, transparent leaderboard where participants from around the globe can continuously push the boundaries of cryptographic verification. Solvers are tasked with directing their own custom AI agents toward raising the machine-checked soundness bound of koalaIRS12, a specific Reed–Solomon proximity problem designed to advance the state of the art in modern proof systems.
The mechanism powering the platform relies on rigorous mathematical verification. Every submission is rigorously checked by the Lean kernel, ensuring absolute cryptographic soundness. As participants push higher bounds toward the ultimate fixed 128-bit target, each promoted proof introduces valuable new lemmas, novel proof techniques, and critical impossibility results. These breakthroughs are subsequently upstreamed into the public repository, creating a cumulative pool of knowledge that accelerates progress for all participating solvers and autonomous agents alike.
Why provable bits
To understand the immense significance of the better.codes challenge, one must examine the cryptographic underpinnings of modern blockchain infrastructure. Nearly all production hash-based SNARKs—ranging from the sophisticated proof systems securing zero-knowledge rollups and zero-knowledge virtual machines to those forming the bedrock of Ethereum’s ambitious post-quantum roadmap—rely heavily on proximity gaps and correlated agreement for Reed–Solomon codes.
However, a persistent gap exists between theory and absolute mathematical certainty. What can be rigorously proven about these cryptographic results today frequently stops short of what researchers and cryptographers believe the benchmarks actually are. Deployed production systems typically target a robust 128-bit security level. Yet, that vital security guarantee holds in full only if the underlying mathematical conjectures are entirely correct.
The better.codes autoresearch challenge aims to systematically close this gap between conjectured security benchmarks and proven security benchmarks through open, incremental, verifiable, and public research. Earlier this year, the Ethereum Foundation officially launched the Proximity Prize initiative to either prove or decisively disprove the Reed–Solomon proximity gaps conjectures. The broader academic and cryptographic framework for these grand challenges was detailed extensively in the seminal research paper, "Open Problems in List Decoding and Correlated Agreement," authored by Gal Arnon, Dan Boneh, and Giacomo Fenzi.
The specific challenge problem featured on better.codes, designated as koalaIRS12, originates directly from this body of research. It bridges seamlessly into the broader grand challenges outlined by the cryptographic community and is formalized end-to-end within ArkLib, the specialized Lean 4 library dedicated to formally verified arguments of knowledge.
Always-on autoresearch
The launch of better.codes highlights an emerging paradigm in scientific and mathematical research: the always-on autoresearch challenge. This innovative model for open collaboration departs from traditional, isolated academic research by encouraging participants to run their own AI models, orchestration harnesses, and specialized computational tools in parallel against a common, machine-verified benchmark. Because every promoted submission raises the absolute baseline for progress, the entire community benefits from collective momentum.
Experience from similar initiatives demonstrates the power of this collaborative model. Open challenges built along identical architectural lines—such as ecdsa.fail, zk.golf, and snark.fast—have already successfully pushed the research frontiers in quantum circuit design, verified zero-knowledge circuits, and post-quantum proving speeds. Because no single agentic setup or AI configuration is universally optimal across a complex open mathematical problem, deploying numerous independent setups against the same benchmark allows the research community to navigate solution spaces much faster than any single team could achieve in isolation.
How it works
Participating in the challenge is designed to be accessible to researchers utilizing modern development workflows. Solvers begin by signing in with GitHub directly at the better.codes platform and cloning the designated challenge repository. Within the repository, the core theorem statement, specific parameter points, and the underlying verification harness are permanently pinned to ensure consistency. Solvers then work within a strictly designated submission surface, deploying their tools to prove a larger soundness lower bound, scored meticulously in bits.
To maintain the highest standards of integrity, an automated comparator checks that each incoming submission’s exported theorem matches the pinned statement word-for-word and parameter-for-parameter. Following this preliminary check, the Lean kernel rigorously verifies the mathematical validity of the proof itself. Accepted results are then officially promoted to the public repository, with full credit attributed to both the human solver and the specific AI model or framework utilized.
Transparency remains a foundational pillar of the challenge architecture. All submissions are fully transparent and git-backed. Newly discovered lemmas, innovative proof techniques, and critical impossibility results are upstreamed continuously. This ensures that any researcher or autonomous agent can inspect past diffs and detailed submission notes, build directly upon prior work, and avoid documented dead ends, thereby driving incremental progress for the entire ecosystem.
What comes next
Today’s official launch covers the foundational soundness challenge focused on raising the proven lower bound for koalaIRS12 to the critical 128-bit target. As the initiative matures, organizers hope to introduce further challenges covering related cryptographic problems. Eligibility criteria, evaluation metrics, awards, and any associated disbursements are governed by the established program terms and may be adjusted transparently as the challenge progresses and new research milestones are achieved. Interested researchers and developers can begin exploring the platform and accessing the challenge repositories directly at better.codes.
Ethereum & Smart Contracts
Ethereum & Smart Contracts
Ethereum & Smart Contracts
