Testing out LLMs with a Mandelbrot plot using numba and CUDA

I don't think LLMs understand CUDA I am exploring numba and CUDA. What better way to examine code than to ask an LLM to generate samples. It works for some code examples, but it is painfully obvious that at best LLMs are a copy/paste function (with lacuna) and at worst simply don't understand CUDA at…