That is true and very useful for software development, but it doesn't help if the goal is to remove human programmers from the loop entirely. If I'm a PM who is trying to get a program to, say, catalogue books according to the Dewey Decimal system for a library, a proof that the program terminates is not going to help that much when the program is mis-categorizing some books.
Is removing the human in the loop really the goal, or is the goal right now to make the human a lot more productive? Because...those are both very different things.
I don't know what the goal for OpenAI or Anthropic really is.
But the context of this thread is the idea that the user daxfohl launched that these companies will, in the next few years, launch an "engineering team replacement" program; and then the user eru claimed that this is indeed more doable in programming than other domains because you can have specs and tests for programs in a way that you can't for, say, an animated movie.
OK, so you successfully argued that replacing the entire engineering team is hard. But you can perhaps still shrink it by 99%. To the point where a sole founder can do the remaining tech role part time.
I have no idea what will happen in a few years, maybe LLM tech will hit a wall and humans will continue to be needed in the loop. But today humans are definitely needed in the loop in some way.
