Post #4417518
2026-08-06 12:24 UTC
@mldkyt@soc.mldkyt.com if possible make it so pieces that result in more interactions get more weight (up to a point)
Replies (1)
-
@hsza@social.tudbut.de 2026-08-06 12:40
@mldkyt@soc.mldkyt.com i can imagine you could: start randomly picking pieces until you have at least n kilobytes of them, take note of what gets picked and run the markov chain on itfor every interaction on the post, add a certain amount of score to every piece involved (fav -> +1, boost -> +3, reply -> +2, or the like)have the random piece selection be weighted by score^0.5 decay score by 1 or 10% (whichever is bigger) for each piece above baseline every iteration