1.3K views
4wk ago
Josh Woodward, Melvin Johnson, Sergey Brin, and Tulsee Doshi from Google DeepMind join host Logan Kilpatrick to introduce Gemini 4 Argon. Their conversation covers what made this release possible and how 100,000 Googlers tested the model before launch. They also discuss industry-leading one million output tokens, improvements to writing quality and model behavior, and the real-world breakthroughs teams are already seeing across coding, cybersecurity, and knowledge work.
Chapters:
0:45 — Welcome and Gemini 4 Argon launch
1:05 — Where Argon is available
1:43 — Early testing with cyber defenders and enterprises
2:50 — Internal testing at scale
3:28 — Building models with models
5:43 — Writing quality improvements and AI slop
8:08 — Agents in the model building process
8:58 — Real world use cases from Googlers
12:50 — How to validate model output
17:36 — What made Gemini 4 Argon possible
18:45 — Aha moments with Argon
23:02 — One million output tokens explained
26:46 — Model behavior, persistence, and transparency
29:00 — Closing thoughts and feedback
🔗Resources:
Gemini 4 Argon Blog → https://goo.gle/4B5pphz
Subscribe to Google for Developers → https://goo.gle/developers
Speakers: Josh Woodward, Melvin Johnson, Sergey Brin, Tulsee Doshi
Products Mentioned: Gemini 4 Argon