I think the future we'll start with unit tests/proofs and then computer will search the solution space find code passes those constraints. The code search will use heuristics.

We'll have to become good at writing those tests, but we won't implement the actual code as much anymore.

I think this the real solution to the "low-code" idea.

Another potential option is simular but the tests will be input/output demonstrations for the computer, and it will figure out the code. It won't be as correct as the first option, but might be good for software that values speed over correctness. Some bad cases may slip through.