Thore Graepel leaves Google DeepMind to pursue AI reasoning venture

Thore Graepel announced on X that he had concluded his latest chapter at Google DeepMind and was turning his attention to what he called one of AI's biggest unsolved problems: reasoning.

Graepel told Sifted that AlphaGo still offers lessons for where AI should go next. His personal site describes the unnamed project as bringing AlphaGo-style reasoning to frontier AI—building systems that can plan and act under uncertainty. He has not disclosed a company name or formal new title.

The move ends a 2025–2026 return to Google DeepMind's Post-AGI team. Graepel previously worked at DeepMind from 2015 to 2021, contributing to AlphaGo, AlphaGo Zero, AlphaZero and MuZero, before leading machine learning for cellular rejuvenation at Altos Labs. He remains Chair of Machine Learning at UCL.