>There is no fundamental principle that suggests most kinds of programming are any less susceptible to automation than any other kind of work.
There are two ways to "automate" programming. Make the language/framework easier to understand/write in, which isn't fundamentally different than it is now, or make the computer understand natural language and do interpretations on its own. The first is likely to happen, but learning to code now will only make you better at writing in whatever simplified language is developed next. The second is AI complete and won't happen till we develop REAL AI... which we aren't really that close to doing.
Learning to code is just learning to be very precise with your instructions. That is a useful skill regardless of whether you are actually coding or not.
What I'm thinking of are a subset of your first thing: powerful task-specific interfaces that act somewhat like DSLs, but usually come with a sophisticated GUI. Some examples: Excel, Photoshop, AutoCAD, Ableton. I think (and hope) these sorts of solutions will proliferate and a bunch of good ones could make a major dent in the world's need for people who program professionally.
I think the "we won't be automated away until REAL AI happens" thing is a trap. Certainly it's true that there will be some programmers working professionally until that point, but I see no reason to expect that it will be a huge number.
I wish I could agree with you more that "learning to code is just learning to be very precise with your instructions". I'm sure there are some programs that stress that, but most of the things I see (especially those targeted at adults) are more focused on the minutia of how we actually write professional software in 2015, because they are targeting job re-training rather than the acquisition of a skill that is broadly useful.
In summary, I'm just not convinced that "learning to code could be a useful insurance policy" is the right message to be sending. Maybe I'd be on board with something more like, "learning to code is probably better than the alternatives", but that leads me to think that we could use some more alternatives.
There are two ways to "automate" programming. Make the language/framework easier to understand/write in, which isn't fundamentally different than it is now, or make the computer understand natural language and do interpretations on its own. The first is likely to happen, but learning to code now will only make you better at writing in whatever simplified language is developed next. The second is AI complete and won't happen till we develop REAL AI... which we aren't really that close to doing.
Learning to code is just learning to be very precise with your instructions. That is a useful skill regardless of whether you are actually coding or not.