I just built SFF RTX 3090 desktop for Llama inference and finetuning (among other things), and I am super stoked about that. More stoked than any PC in a long time.

Its not productive, but it's fun. The whole generative AI field is like a big sandbox full of magic, janky crystal balls.