"There's also the CUDA moat which is hard to get around"
Is it really, though? Porting to a new architecture is an ideal use case for LLMs.
clearly they did not read the article
clearly they did not read the article