Unfortunately, this is simply not the case, as much as I hate it. I've worked on several custom ISAs and processors (mostly at companies shipping custom AI acce

Unfortunately, this is simply not the case, as much as I hate it. I've worked on several custom ISAs and processors (mostly at companies shipping custom AI accelerators), and after sufficient explanation of the ISA to chat gpt, it was able to write machine learning kernels that did indeed run on our architecture. It is an impressive piece of technology, as much as I hate to admit