[tslug]

Yes, in general, available work will decline significantly as will pay. That said, there will be an ever-narrowing niche for work that requires extremely high skill and that pays unprecedentedly and increasingly high returns. This is because every time a piece of software antiquates a former programming market, that software requires a shrinking group of even more talented coders to develop.

We're seeing this in machine learning right now. Machine learning can replace a lot of the code we used to implement with expert systems or rote heuristics. If you can develop that code (and not many can), you can command unusually high salaries, but there will come a time when developing machine intelligence becomes fairly straightforward itself. In fact, this is already beginning to happen. And then even these coders with their rare skills will start to see their salaries fall.

This has been a consistent pattern throughout the history of software engineering. However, I believe we're going to see salaries fall off a cliff soon, because I believe we're actually quite close to automating good software engineering in general.

It's going to be amusing, as a lot of the best coders feel particularly superior and entitled to their impressive compensations.

[whenwillitstop]

Yeah I think the same thing. If we compare Hadoop to Spark, Spark is just easier and faster. It is superior in almost every way (unless we are talking about 1PB+ scale). I think data science will be automated. Some things wont, but engineers can already get pretty far by jamming data into ML packages.

[mgberlin]

"I believe we're actually quite close to automating good software engineering in general."

What leads you to believe this is true? At the moment we don't even have good automation for telling us where a bug is in a stack trace, or a solid definition of "good software engineering."