every word is one step left-to-right; the line climbs by one each time a word
arrives that hasn't appeared before, and holds flat on a
repeat. so the curve is the text's vocabulary filling up in reading-time — a
freshet of new words at the start, thinning to a
trickle of echoes. the faint diagonal is the
if-every-word-were-new ceiling; the gap below it, growing rightward, is
everything the text has already said. the one tell: where
the climb peels away from the diagonal is where the writer stops introducing and
starts returning. every text is concave — that's Heaps' law, vocabulary always
saturates — so flattening is universal and says nothing; the signal is the
rate, and where the fresh words cluster. a vivid passage steps
the line up in a burst; a refrain lays it flat. kin to
hyperloglog (which counts the distinct total
and pays nothing to hold it) and hapax (the words
used exactly once) — but those report a number; freshet draws the order the
number was earned.