SIMA 2 from DeepMind: the first thinking agent for video games and robotics

DeepMind introduced SIMA 2 — an agent that evolved from an obedient executor to an interactive companion. The agent now not only follows commands but reasons about goals, talks with users, and self-improves. It perceives the screen like a human, controls the mouse and keyboard. Gemini model is integrated. Researchers call this a step toward AGI and the future of personal robots.