Show HN: ZkGolf

Zero-knowledge proofs are notoriously difficult to implement — they require circuits, essentially cryptographic assembly code that's heavily hand-optimized. According to Hacker News, researchers have discovered that language models can actually do this better. They created ZkGolf, a competition for optimized, formally verified zero-knowledge circuits. The breakthrough came when a large language model generated an S-H-A two fifty-six circuit that not only worked, but beat current human-optimized versions. The model could even optimize automatically, proving each change was mathematically sound and backtracking when it wasn't. The goal is to lower the barrier to entry for zero-knowledge proofs by making them cheaper and faster to produce.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton