PodcastIntel
Sign in Get Started Free
Neural intel Pod
Neural intel Pod

Inside the Machine: Training GPT-5, the Memory Wall, and the Math of MoE

May 1, 2026 · 00:45:18
AI Summary
  • Discusses training and serving advanced models like GPT-5 and Gemini at scale.
  • Analyzes hardware constants dictating optimal batch sizes and sparsity ratios.
  • Explains the 20ms rule and its implications for memory capacity.

Guests on This Episode

RP
Reiner Pope
3 podcast appearances

More from Neural intel Pod

Neural intel Pod
Jul 7, 2026 · 00:40:29
View all episodes →

Get AI Summaries for Every New Episode

Subscribe to Neural intel Pod and get AI summaries, guest tracking, and email digests delivered automatically.

Sign Up Free →