Tretij Rebenok

X’s algorithm has a ranking weight of 20.0 for something almost nobody talks about. It’s NOT likes. It’s NOT reposts. It’s NOT replies. It’s the predicted likelihood that someone COPIES YOUR POST’S LINK. The X’s open-source algorithm should change how you create content. Here’s what the actual code says: 1. These are X’s published ranking weights: • Copy-link share: +20.0 • Reply: +5.0 • Quote post: +5.0 • DM share: +5.0 • Follow: +4.0 • Repost: +1.0 • Like: +0.6 Read that again. The coefficient for a predicted copy-link share is over 33x the coefficient for a predicted like. But there’s a catch. These aren’t points awarded for actual interactions. X multiplies these weights by the probability that a particular user will take each action. 2. X isn’t simply counting engagement. It’s predicting YOUR behavior. For every post, the system asks: → Will this person reply? → Will they share it privately? → Will they spend time reading? → Will they follow the author? → Will they mute or block them? Those predictions help determine what appears in your For You feed. 3. Negative feedback matters too. The code includes these negative weights: • Not interested: -47.52 • Mute: -58.8 • Block: -31.2 • Report: -234 Again, these affect predicted probabilities, not raw engagement counts. Manufacturing outrage might generate replies. But it can also make people want to never see you again. 4. Here’s how I’d actually use this information. Stop asking: “Will people like this?” Start asking: “Would someone send this to a friend?” For SEO creators, that means publishing: • Actual traffic experiments • Before-and-after GSC data • Useful workflows people can replicate • Original research worth referencing • Contrarian opinions backed by evidence • Frameworks worth sharing with a colleague Give readers a reason to share your work privately, quote it publicly, or follow you for more. 5. Your audience matters more than vanity metrics. X combines posts from accounts people follow with content discovered through machine learning. It then ranks candidates for individual viewers. You don’t need every person on X to like your content. You need the RIGHT people to find it valuable. The biggest lesson from the code? Stop chasing likes. Start creating things people want to share, discuss, reference, and come back to. You can’t guarantee virality. But you can give the algorithm better reasons to recommend your content. What’s the last post you shared with someone privately? That’s the kind of content I’d study. Source: github.com/xai-org/x-algorit…