FORTRAN was also going to end programming: IBM said so in 1954

The essay: eight decades of "programming is over"

The essay treats AI as a faster FORTRAN. Its evidence is about review. Faros AI data showed more pull requests but much longer review times, and some teams merged unreviewed code. A METR trial found experienced developers took 19% longer with AI while believing they were quicker. Agent task length has roughly doubled every six months, though the author calls that a trend, not a guarantee. A Stanford study found employment of software developers aged 22 to 25 down nearly 20% from its late-2022 peak, with unclear causes.

The author predicts smaller teams, specs and tests as the primary artifacts, "verification" roles, and labels showing who or what wrote each change. The hard part, deciding what should exist and specifying it, has always been the job.

The thread: is a compiler even the right comparison?

"With compilers there is a (roughly, barring optimizations) 1:1 correspondence between input and output. With LLMs you take an idea and it will drive the whole thing from the beginning all the way to the end." — Jacques Mattheij

"However, the human language is not a programming language." — a commenter

A third reader went after the history.

"Nope; the compiler was supposed to end coding; when the word coding referred to taking an abstract program written on paper and translating it to machine code. Compilers were first called automatic coders." — kazinator, who works on the TXR language

By that reading the 1954 prediction came true: compilers did end hand translation to machine code, and what remained was the part that needed judgment.