DeepMind veteran Thore Graepel has left the company to launch a new venture focused on AI reasoning. Graepel, who joined DeepMind in 2016, was a key figure in the development of AlphaGo, the program that defeated grandmaster Lee Sedol in 2016. His departure marks another significant exit from the Google-owned AI lab.

Graepel confirmed the move on LinkedIn, stating he is pursuing a "new adventure" centered on reasoning capabilities, a critical frontier for achieving artificial general intelligence. He expressed pride in his work at DeepMind, including contributions to AlphaFold and large language models. The startup is currently in stealth mode, with details on funding and specific technical approach scarce.

This move follows a trend of top AI researchers leaving major labs to found reasoning-focused startups, highlighting the intense competition and investment in the sector.