Google lost two of its most important AI figures in a single day, and Wall Street noticed immediately — Alphabet shares slid 5.4% intraday on August 5.
Jeff Dean, known inside Google as employee number 30 since he joined back in 1999, is stepping down as chief scientist after nearly 27 years. His fingerprints are on MapReduce, Bigtable and TensorFlow, and in 2011 he co-founded Google Brain, the lab that became the backbone of the company's entire AI strategy.
Dean isn't jumping to a rival — he's starting his own company, Discovery Loop, alongside three other Google veterans: engineering fellow Sanjay Ghemawat, Google Brain co-founder Quoc Le, and DeepMind researcher Oriol Vinyals. The pitch is bold: automate science itself. Discovery Loop wants to run thousands of experiments in parallel and eventually push toward what the founders call recursive self-improvement, AI helping build better AI with less human involvement at every step. Radical Ventures and Khosla Ventures co-led the seed round, with Kleiner Perkins, Lightspeed and Doerr Capital also participating — and Alphabet itself joined as a founding investor and cloud partner.
DeepMind's corner office changed hands too
On the same day, Demis Hassabis handed off his day-to-day duties as CEO of Google DeepMind. He stays on as chair of the division and picks up a newly created title, chief scientist of Alphabet, while continuing to run Isomorphic Labs, his drug-discovery venture. Day-to-day control of DeepMind now goes to Koray Kavukcuoglu, who has worked alongside Hassabis for 13 years and previously served as chief technology officer. He now reports directly to Sundar Pichai and oversees Gemini model development, frontier research and the Gemini app.
It's not the first high-profile exit for Google this year — Noam Shazeer left for OpenAI and John Jumper departed for Anthropic earlier in 2026. Gemini remains competitive on the leaderboards, but keeping the people who build it is proving far harder than keeping the models themselves at the top — and that gap, more than any single benchmark score, is quickly becoming Google's biggest liability in the AI race.
