Every Non-Word is a twitterbot that attempts to exhaust all possibilities of non-existent English words. It's been running since March 2015 and was a featured project for the 2017 edition of The Wrong, at which point it had tweeted 36,000 non-words.
Each non-word is a randomized collection of syllables, first collected by running a hyphenation script against the entire English dictionary. When they were loaded into a database, the doubles were not removed, so that when syllables are pulled, they are already weighted by how often they appear in English. Each resulting word is checked against a dictionary to rule out accidental real words; if it does match something, the bot tries again.
Freed from denotative meaning, the resulting words are clouds of semantic possibilities and associations. Strange compound words sometimes result, appearing more like several complete English words stuffed together. These are often the most retweeted: wetmood, automoon, and the endlessly popular endhetero. However, the more vague words are perhaps the most interesting. While the bot is built on English composite parts, Twitter's built-in Bing translater sometimes identifies Every Non-Word tweets as a variety of other languages; Hindi, Swahili, Czech, many others. Sadly, when one clicks for the translation, they result in the "Could not translate tweet" message.
Every Non-Word was named for Allison Parrish's Everyword, the Twitter bot that listed every English word in alphabetical order, completing in 2014. While Allison Parrish's bot had a limited run by design, @nondenotative's corpus, while not endless, will still likely be growing when Twitter finally shuts down. If the bot were to actually exhaust all possibilities, it will not crash, but will enter an endless loop of crafting random words and rejecting them as being English or previously tweeted.
A series of 100 zines produced with the Every Non-Word algorithm, each with a unique set of words.