The Pomp Podcast Inside Google's Billion Dollar Bet To Win The AI Race | Logan Kilpatrick
63 snips
Sep 15, 2026 
Logan Kilpatrick, a Google DeepMind technical staff member focused on frontier AI and Kaggle, explores Google’s bid to reclaim the AI lead. They discuss Gemini 4’s billion-dollar-scale training run, buying versus building coding expertise, China’s open-weight labs, and the tension between general intelligence and specialized products. They also examine AI benchmarks, DeepMind’s science flywheel, and why data may become the next bottleneck.
AI Snips
Chapters
Transcript
Episode notes
Google Uses Acquisitions To Accelerate Internal Coding
- Google has pursued coding through both internal development and acquiring the Windsurf team.
- The acquired team’s internal Anti-Gravity product reportedly accelerated Google engineers before becoming an external offering.
Gemini 4 Signals Google’s Frontier Commitment
- Gemini 4’s massive pre-training run signals that Google is still willing to make frontier-scale investments.
- Google’s TPU fleet, data infrastructure, and pre-training expertise provide structural advantages in this effort.
Why Open-Weight Models Matter To Startups
- Chinese open-weight labs are serious competitors because they combine research innovation with models customers can host and customize.
- Startups increasingly value control over model weights, creating a major market beyond closed APIs.