Programming languages

A programming language is a notation for expressing computation that people can read and write, paired with a translator that converts it into the numeric instructions a processor executes. Instead of writing the literal opcodes a machine understands — long strings of numbers specifying registers and addresses — you write `AREA = PI * R**2`, and a compiler works out the exact sequence of machine instructions that computes it. The language defines what you are allowed to say (its syntax and semantics); the compiler or interpreter is the machinery that faithfully turns what you said into what the hardware does.

The earliest stored-program computers were programmed in raw machine code or, slightly better, assembly language — a painstaking craft understood by a small "priesthood" who thought in the machine's own numbers. That was the bottleneck: programs were slow to write, riddled with errors, and locked to one model of computer. High-level languages broke it. IBM's FORTRAN, whose first compiler shipped for the IBM 704 in 1957 under John Backus, let a scientist write ordinary algebraic formulas and have them run nearly as fast as hand-tuned machine code — the objection that mattered most at the time. Suddenly people who were not machine specialists could program, and a program written in the language could be moved to any computer that had a compiler for it.

Programming languages are the reason software is an industry rather than a specialty. Layer upon layer of abstraction — assemblers, then FORTRAN and COBOL, then thousands of languages since — let programmers describe ever larger systems without tracking the machine's every register, and let the same program run on wildly different hardware. Every application, operating system, website, and AI model is written in one language and compiled or interpreted down toward another. The whole pyramid rests on the modest act of translation Hopper and Backus made routine: say it in human terms, and let a program turn it into the machine's.