Скачать или смотреть How to choose local LLM AI Hardware (Pattern Practice Podcast 125)

Cyril and Erik join to pitch a family LLM machine—local, private, expandable. What starts with frustrations over API keys and privacy pivots into building a Raspberry Pi trash classifier. From there the crew riffs on GPUs versus Apple Silicon, stacking Mac Minis, and the sweet spot of 24 gigs of VRAM. Cyril floats the idea that his rig might pay for itself by selling idle cycles, while Joe and Josh nudge toward resale value and agent use cases. The conversation closes with validation, Mac Mini endorsement, and the promise to pick up with node-based automation tools next time.

00:01 Family LLM Machine
02:56 Sort by Trash
11:31 Stacking Mac Minis
31:29 Pay for Itself and Agents on the To-Do List
46:37 Next Session: N8N and Wrap-Up