Google's Stellar Colosseum reports a 4263 Codeforces rating with a many-agent proof harness.

arXiv 2609.15983 from Honghao Lin, David P. Woodruff, Vahab Mirrokni and colleagues explores alternative proof strategies in parallel, uses a readiness gate before decomposing a plan into interdependent subproblems, and routes verifier feedback back to the specific failing part of the argument. It solves 218 of 222 Codeforces problems for 4263 against a 4039 best human score, and reaches 71.0% on TCS-Bench using Gemini 3.1 Pro and Gemini 3.7 Flash. arXiv The harness has been folded into Antigravity's Teamwork framework.