Season 2 · Episode 1931
AI Pipelines: In-Memory vs. Durable State
Why do AI pipelines crash? It’s not the models—it’s the plumbing. We break down how to manage data between stages.
My Weird Prompts · Daniel Rosehill
April 2, 202623m 14s
Audio is streamed directly from the publisher (dts.podtrac.com) as published in their RSS feed. Play Podcasts does not host this file. Rights-holders can request removal through the copyright & takedown page.
Show Notes
Everyone obsesses over frontier models and prompt engineering, but production AI fails at a more fundamental layer: the plumbing. This episode dives into the unglamorous but critical world of state management in multi-stage AI pipelines. We explore the trade-offs between volatile in-memory passing, high-speed caches like Redis, and durable databases, and introduce frameworks like LangGraph and Temporal that promise "immortal" execution. Learn why the "where" and "how" of data movement determines whether your system is a brittle prototype or a resilient enterprise tool.