ROT13 was never invented by a cryptographer — it evolved on the
early internet. In the 1980s, the Usenet discussion system needed a
way to post potentially offensive jokes and film or book spoilers
without ruining them for scrolling readers: the text should be
present but not accidentally readable. A fixed Caesar shift of 13
was the obvious answer, and the newsgroup net.jokes made it a
convention.
Two reasons, both practical. First, 13 is self-inverse — the same procedure hides and reveals, which meant readers could use one command (or one glance at a printed wheel) for both directions. Second, 13 moves every letter off its original: with an even alphabet, only the half-turn pairs every letter with a different partner, so no letter betrays itself by staying put.
Through the 1980s and 1990s ROT13 became part of netiquette: spoiler
warnings, puzzle hints, punchlines, even answers to quiz posts. Unix
grew a rot13 command; newsreaders built it in. Security
professionals cite it as the canonical example that obfuscation is
not encryption — and puzzle-setters keep it canon for exactly the
same reason.
Decades later the wheel keeps spinning: certificate transparency logs hide hostnames in ROT13, geocaching puzzles use it, and every new programmer writes their own ROT13 as an early exercise. The oldest joke-hider on the internet is also the longest-lived.