Subvolts take Google's Gemini leaders reveal project origins and future plans for their unified AI models. Jun 22, 2026 - Developers
- Learn about the strategic decisions behind Google's Gemini models and their performance improvements.
- Business
- Understand Google's approach to developing large AI models and their future roadmap.
This episode features key leaders discussing the foundational strategy and future direction of Google's flagship AI models.
Gemini co-leads on project origins and what's next
Audio streams from Google AI. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Google AI: Release Notes features Logan Kilpatrick interviewing Gemini co-leads Jeff Dean, Koray Kavukcuoglu, Noam Shazeer, and Oriol Vinyals. They discuss the Gemini project's origins, the strategy behind a unified model, the performance improvements in Gemini 3.5 Flash over Pro, and future developments. The conversation also touches on the motivation for creating Gemini 3.5 Flash.
Google's top AI builders discuss how they created their Gemini AI. They explain why they built one big model instead of many small ones and share what's next for Gemini's abilities.
This episode offers insight into the strategic decisions and technical evolution behind Google's Gemini models, particularly the new 3.5 Flash. It's for AI researchers, developers, and tech leaders interested in Google's LLM development. The discussion on unified models and performance gains connects directly to the ongoing frontier lab race, showing Google's approach to competing with models from OpenAI, Anthropic, and Meta.
- HostLogan Kilpatrick
- GuestJeff Deanco-lead, Gemini
- GuestKoray Kavukcuogluco-lead, Gemini
- GuestNoam Shazeerco-lead, Gemini
- GuestOriol Vinyalsco-lead, Gemini
- Episode Title: Gemini co-leads on project origins and what's next
- Show: Google AI: Release Notes
- Length: 42 minutes
- Guests: Jeff Dean, Koray Kavukcuoglu, Noam Shazeer, Oriol Vinyals
- Topics Discussed: Gemini project origins, bet on a single unified model, Gemini 3.5 Flash performance vs Pro, what's coming next
- unified model
- Koray Kavukcuoglu
- building one large AI model that can do many tasks, rather than many specialized ones.
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
39 minIntroducing Gemini Robotics 2
WithCarolina Parada, Stuart Bowers, Kanishka Rao, Jie Tan
40 min1 h 12Yossi Matias on the golden age of research
WithYossi Matias
Same guests, other shows
27 minKoray Kavukcuoglu on frontier models, coding agents, and building AGI
WithKoray Kavukcuoglu
1 h49 minKoray Kavukcuoglu: “This Is How We Are Going to Build AGI”
WithKoray Kavukcuoglu
49 minGemini 2.0 and the Evolution of Agentic AI with Oriol Vinyals
WithOriol Vinyals
50 minLiz Reid on What Publishers Get Wrong About AI
WithLiz Reid
2 h 27The $10B Satellite Empire Putting AI in Orbit, Why Chips Beat Rockets & China's #1 Open Model | EP #266
WithWill Marshall, Salim Ismail, Dave Blundin, Alexander Wissner-Gross
Questions people ask
- Who is interviewed in this episode?
- Logan Kilpatrick interviews four Gemini co-leads: Jeff Dean, Koray Kavukcuoglu, Noam Shazeer, and Oriol Vinyals.
- What is discussed regarding Gemini 3.5 Flash?
- The episode covers the origin of the Gemini project, the bet on a single unified model, why each Flash generation outperforms the previous Pro, and what's coming next.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Google AI.