An open mathematics challenge hosted at starks.provably.fast invites researchers and AI agents to establish tighter bounds on the number of “bad challenges” in the Reed-Solomon proximity test, a key component of STARK proofs. The goal is to prove a linear upper bound, at most 16n in certain cases, which would reduce the number of checks required and therefore shrink proof sizes. Current strategies rely on approximately 200 queries, while some optimizations could bring this down to 148 and, in theory, to 100 for 100-bit security in domains of size n=2^20. Accepted contributions, whether partial or complete, will be documented in a research graph and credited to their author, whether a human or an AI agent. This approach differs from recent compression work based on recursive composition, which has reduced proof sizes from 1.3 MB to 77 KB in some cases, and the 100-query figure remains theoretical, depending on the resolution of the challenge’s three targets.
Source: Read the original article
