Koray Kavukcuoglu on frontier models, coding agents, and building AGI
Summary
Google DeepMind's Koray Kavukcuoglu discusses building AGI, highlighting Gemini 3.7 Flash as a breakthrough and the ambitions behind the Gemini 4 pre-training run. He explains that co-building AGI with users shapes the team's work and emphasizes the importance of staying at the AI research frontier. Kavukcuoglu also touches on the journey from early reinforcement learning to current models and the challenges of engineering at this scale.
Why it matters
This video offers a glimpse into Google's AGI ambitions, positioning Gemini as a key component. Kavukcuoglu's insights into Gemini 4's pre-training and the concept of 'co-building AGI with users' suggest a user-centric development approach. The discussion on 'engineering hill climbs' and the lack of an AGI test highlights the ongoing, complex nature of frontier AI research. This contrasts with competitors who might focus on specific benchmarks or product integrations, indicating Google's long-term, foundational research focus.
Related: OpenAI: Sam Altman :‘AGI in 2026’, just as Models Start to [Mis]Train Themselves · Anthropic: Claude Fable AI Is Much Stranger Than The Headlines Suggest · Microsoft: Orchard: An open framework for scalable agentic AI · Meta: An Organizational Second Brain: Building an AI That Learns From Experts · xAI: Ajeya Cotra – "This might be the clearest warning shot we ever get"
Rated middle: a real update, not a headline event.
- Google’s latest AI weather model gives you no excuse to forget your umbrellaTechCrunch AI · Web · Sep 3, 2026
- ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide GenerationarXiv cs.AI · Web · Aug 31, 2026
- Four major AI models suffer rare overlapping downtimeArs Technica AI · Web · Sep 3, 2026
- Four major AI models suffer rare overlapping downtimeArs Technica AI · Web · Sep 3, 2026
- Black Box: The Chatbots | 14 days | Ep 2 – podcastThe Guardian AI · Web · Sep 3, 2026
- Black Box: The Chatbots | Spirals | Ep 1 – podcastThe Guardian AI · Web · Sep 3, 2026
- ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide GenerationarXiv cs.AI
- Dr. Claw: An AI Scientist Workspace for Vibe ResearcharXiv cs.AI
- 7 AI agent patterns to improve your coding workflowGoogle Cloud Tech on YouTube
- Piloting the world's first double-blind AI evaluationsGoogle DeepMind Blog
- Pairing Google Antigravity with Gemini 3.7 Flash solves notable multi-agent math and engineering problems.The Keyword: Developers
- Launch your to-do web app with AIGoogle Cloud Tech on YouTube
Questions people ask
- What made Gemini 3.7 Flash a breakthrough?
- The video states Gemini 3.7 Flash was a breakthrough for engineers, though specific details are not provided in the source text.
- What is Google's approach to building AGI?
- Google DeepMind builds AGI by staying at the research frontier and co-building AGI with users, which shapes their development process.
- Why is there no test for AGI?
- The video suggests there is no definitive test for AGI, implying its complexity and the evolving nature of intelligence.
More from Google for Developers on YouTube 14 more
Build a live translation broadcast app with the Gemini Live API and LiveKit
Build a live translation broadcast app with the Gemini Live API and LiveKit
How builders at YC Startup School are using Gemini & Google AI
Page generated Sep 3, 2026. Summaries are Subvolts' own; the story belongs to Google for Developers on YouTube.