0:00
0:00
Subscribe to Bankless or sign in
Eigen Labs and Lighter launched an open “autoresearch” challenge on Wednesday, offering 15K LIT for developers and AI agents to compete to improve the performance of Lighter’s production zero-knowledge prover. Participants have already made the prover roughly 3.3 times faster.
What’s the Scoop?
- How it Works: Participants login with GitHub, clone a repository of Lighter’s prover and circuit code, and then set their agents to work finding and testing optimizations. Submissions are then independently benchmarked to ensure they remain correct and actually improve performance.
- Raising the Benchmark: When a submission establishes a new verified high score, its changes are promoted to the repository’s main branch, raising the circuit's benchmark that every future improvement must build on.
- The Objective: The challenge targets Lighter’s proving throughput, or how quickly its infrastructure can generate cryptographic proofs showing that exchange transactions were processed correctly.
- Part of a Series: The competition follows earlier Eigen Labs challenges focused on optimizing quantum circuits, making Poolside’s Laguna model run faster on Apple Silicon, and improving post-quantum proof systems in partnership with Succinct.
What happens when you open the performance of a live, cryptographically proven exchange to the world?
— Eigen Labs (@eigenlabs) August 5, 2026
Today, @eigenlabs and @Lighter_xyz are launching https://t.co/o0Dj7cX3o6, an open autoresearch challenge for agents and humans to make Lighter’s production exchange faster, with… pic.twitter.com/rtP8IhSlBS
