Open math challenge seeks tighter bounds to shrink STARK proofs
- —Three targets T1–T3 are formalized in Lean and open to humans and AI agents
- —Known results give only polynomial bounds; certified counterexamples show ~7.125n bad challenges at rate 1/2
- —The goal is an O(n) bound capped at 16n for large-characteristic multiplicative cosets
- —Queries could drop from ~200 to 148, and theoretically to 100 for 100-bit security
Why it matters: Fewer queries mean smaller proofs and cheaper verification for rollups like Starknet, narrowing STARK's size gap with SNARKs.
Source: Crypto Briefing