Comment Re: Well "just" vibe code you a new API, then eh? (Score 1) 46
I know everyone makes the same joke regarding AI not being able to do it. And whoever works at the edge of any field knows that the sort of large projects required to be implemented cannot simply be asked a swarm of agents to code, some things are, and will stay out of reach for a while, maybe forever.
But, could any generous soul explain some specifics about the sort of things that make it that hard? What are the challenges there? Is it the amount of code required to be translated? The secrecy of the internals of CUDA? Is it how undeveloped the other platform is? Or maybe how different the abstractions are on each side? Or maybe something I have not though about. Please explain for those of use that don't do any GPU coding.
But, could any generous soul explain some specifics about the sort of things that make it that hard? What are the challenges there? Is it the amount of code required to be translated? The secrecy of the internals of CUDA? Is it how undeveloped the other platform is? Or maybe how different the abstractions are on each side? Or maybe something I have not though about. Please explain for those of use that don't do any GPU coding.