DayNews.ai

AI agents now tested as autonomous researchers

A new benchmark tests AI coding agents by having them independently improve world models, not just follow instructions.