Oriol Vinyals

Google Gemini Deep Think and OpenAI's experimental reasoning model both achieved International Math Olympiad Gold medal performance by solving 5 of 6 problems within human time limits using general-purpose reinforcement learning and test-time compute scaling.

“Commented on Google DeepMind's achievement of IMO Gold with Gemini Deep Think using pure in-weights reasoning without specialized tools.”

Elon Musk proposes Grok 5 AI compete against the world's best League of Legends team under human-level constraints, with positive reception from professional players and Riot Games suggesting the exhibition match could materialize.

“AI researcher who previously reached Grandmaster rank in StarCraft and suggested testing Grok in RTS gameplay.”

Gemini 3's breakthrough performance improvements with identical parameter count to Gemini 2.5, combined with Nvidia's Blackwell infrastructure dominance, proves scaling laws remain viable and the 2025 'scaling wall' narrative was unfounded.

“Credited improving pre-training and post-training for Gemini 3's massive performance improvements, stating the delta between 2.5 and 3.0 is the biggest Google has ever seen with no scaling walls in sight.”