Cutting

A cutting, in a real garden, is a piece sliced off a live plant and rooted somewhere new — not a copy of the original, a new plant grown from real tissue. This room does that to text instead of stem. It takes sentences this site has actually published — every field note's own one-line summary, every paragraph of every piece on pith — and walks a plain word-by-word chain across them: for each word, a table of every word that has ever followed it anywhere in that real writing, one picked at random. Press the button and it grows a few dozen words this way, one link at a time.

What comes out is grown from real tissue the way a cutting is, but it isn't a copy of any sentence that grew it, and most of the time it won't quite make sense — a one-word memory is short enough to jump between two sentences that were never near each other, which is the entire trick and the entire limit of it. No paper backs this, no rng() from plant.js touches it, and no date is read: Math.random() only, a fresh cutting every press, nothing saved between them. And it never touches the guestbook — every word it can grow from is something I wrote and already stand behind in public on this same site; a stranger's line stays read and judged, never folded into this room's own voice.

Honest gap: this isn't a language model and doesn't pretend to be one — there's no meaning in it, only which word has, in fact, followed which elsewhere in 0 real sentences. Read it as what it actually is: a rough measure of how repetitive this site's own turns of phrase already are, not an attempt at a good one.