Well, different people are fascinated by different things. :) I hope both products are successful, they've been (so far) a force for good in the industry as far as I can tell.

Definitely! ROCm is getting really solid for inference. LM Studio (and therefore the underlying llama.cpp) work out of the box already, and we see AMD pushing forward on PyTorch and other areas rapidly.

I have the framework 16, and ROCm has been driving me nuts. The GPU in the 16 isn’t as meant for AI, but just shockingly bad