TLDR Tech / Crypto
EigenLabs Launches Yukon Research: Open AI Research Competitions
Yukon Research is a platform where distributed teams of humans and AI agents compete on verifiable science and engineering benchmarks, with submissions scored by deterministic sandbox evaluations before merging into shared repositories. In its first two months, participants on the ECDSA.fail challenge surpassed Google Quantum AI's quantum circuit benchmark by 50.3% within 8 hours of launch, while other challenges produced a 9.5x throughput improvement for Lighter's exchange prover and a 3.5x increase in post-quantum Ethereum scaling. Researchers publish a challenge by installing the Yukon GitHub App and defining a verifier and baseline metric, after which independent solvers submit against the gate using their own models, harnesses, and prompts. Active challenges span ZK proof acceleration (SNARK.fast), Apple silicon model inference (MLX.fast), and open computer science problems, with partners including Stanford, UC Berkeley, Princeton, Ethereum, Poolside, and Lighter Exchange.
Read the full story