Former DeepMind researcher founds Metis Reasoning to tackle hallucinations
Understand the story
On October 27, 2026, former DeepMind researcher Thore Graepel announced the founding of Metis Reasoning at the NVIDIA GTC conference in Berlin. The company proposes bringing AlphaGo's search methods into general reasoning to reduce hallucinations in large models. Its technical approach includes: using an explicit knowledge ledger to record known information, confidence levels and open questions, with an independent module verifying each reasoning step. Graepel says this method supports traceable, deletable knowledge management. The company is currently at the startup and technical-proposal stage, with no product launch, funding or further news yet.
AI overview of reporting · Updated 1 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderFormer Deepmind researcher thinks AlphaGo's playbook can cure AI hallucinations
Former DeepMind researcher Thore Graepel has founded Metis Reasoning, proposing at the Nvidia GTC conference in Berlin to bring AlphaGo-style search methods into general reasoning, using an explicit knowledge ledger that records what is known, confidence levels and open questions, with independent modules verifying each reasoning step to reduce hallucinations and support traceable, deletable knowledge.
Story heat trend
Not enough continuous observations to show a trend yet.