Makes sense to me. Until we get diminishing returns from LLMs it seems like we're on track to get something like that sooner rather than later. I mean at some p

Makes sense to me. Until we get diminishing returns from LLMs it seems like we're on track to get something like that sooner rather than later. I mean at some point if we can make an AI that actually understands computer science, software engineering, and human intentions well enough then why have compilers? The AI itself will just "compile" from human language to machine code. Humans will learn to just double-check its work the same way that non-programmers currently double-check programmers and the same way I double-check compilers in my day job.