OpenAI dropped EVMbench, a benchmark to evaluate agentic AI's ability to detect, patch, and exploit vulnerabilities in smartcontracts announcement: https://openai.com/index/introducing-evmbench/ tool: https://github.com/openai/frontier-evals