reads every recent post from you and everyone in your SimCluster, then runs it through ADVANCED ALGORITHMS (vocabulary, hashtags, posting style, time-of-day rhythm) to find your closest match — your twin — and your most opposite match — your anti-twin.
every account's recent posts get turned into a word-frequency vector
(stopwords stripped), a hashtag set, an average post length / reply
ratio / emoji rate, and a 24-bucket UTC hour-of-day histogram. your
vector gets compared against each mutual's with cosine similarity on
vocabulary and rhythm, Jaccard overlap on hashtags, and a normalized
distance on posting style — weighted average of whichever of those
four have enough data, 0–100. highest score is your twin, lowest is
your anti-twin. it's real math. it is not a very deep model — two
people who both post a lot about the same show will out-score two
people who share a soul but never say #bluesky.