I believe the discussion should go beyond the question of 'Programmers Replacement'.
Lets assume that the numbers of amount of code written today by AI are a trajectory rather than an assumption.

Lets assume AI will replace human programmers, in most, if not all programming positions.

The question now should evolve to the programming languages themselves. Because if this is the future, than we are holding it back with programming languages that are bound by human limitation needs.

Why does AI needs to write Python? or any other human-fitted language for that matter?

Roughly 97% of what programmers write is not computation. It is communication to machines and other humans that may read their code.

When code no longer needs to be read by human programmers there is -

• No need for meaningful variable and function names
• No need for whitespace, formatting, boilerplates, imports, class declarations, constructors, getters, setters, comments, documentation
• There will be no Syntax Errors
• No software stack

Code becomes 20x - 50x more effective and humans remain to deal with the 'What" and not the "How".

It becomes clear that those who can best articulate the intent and specification of software (rather than its code) — and be instrumental in narrowing said Intent Dissonance Gap — will become the future social computing artists in high demand.

I have written a detailed essay on this as I believe this is the next area of evolution that needs discussion.