How close do the in-browser embeddings actually get to what the real pipeline produces? Or is it more about showing the general flow than matching real outputs?
prabal97•1h ago
They're weighted differently. Currently it's set to: reposts 2.0x, replies 1.5x, bookmarks 1.2x, likes 1.0x, and clicks 0.5x. These are guestimates based on public info.
blasphemous_dev•1h ago
Is the scoring model actually weighting engagement actions differently or is it treating likes/replies/reposts the same? The real algorithm weights those pretty aggressively differently
prabal97•1h ago
It's real embeddings (MiniLM, truncated to 128-dim), so the cosine similarities between tweets are genuine, but it's obviously not X's actual trained model. The relative distances are meaningful, though.
saranshmahajan•1h ago
prabal97•1h ago