The number of machinists in the United States has decreased by about 3% over the past decade.

During this time, computer-numeric-control (CNC) machines have been adapted for many manufacturing processes that previously might have called for a machinist.

Naively, you might assume that this change would drastically reduce the number of machinists. Instead, much more efficient creation of machines and equipment seems to provide its own offsetting need for machinists (to run the CNC mills and assist with maintenance).

I think this is a likely model for generative language models. Yes, code may become much easier to generate, but more voluminous code requires more maintenance and code generation is not so trivial as to require zero human input, skill or training. Critically, it is already the case that maintenance consumes far more coding time in most projects than does new code generation. Making new code generation easier exacerbates this and should be expected to *increase* the demand for trained computer programmers.