Innocent people don’t plead the Fifth on basic questions. Yet somehow our tax dollars are paying for this circus.

Until Monday, the X algorithm didn’t factor in who your mutuals were. That data was missing from the model. I give credit to X for saying it out loud and then shipping the fix the same day. Transparency matters. A lot. Now that we’re openly talking about improving the algorithm, I have a suggestion that moves us further in the same direction. I’m not a hater. I’m here every day. I want this place to win. I brought a use case for consideration: THE USE CASE • Same account (mine) • Two posts • Same followers • Hours apart • One got 7,700 reposts • The other got 136 A controlled experiment nobody designed. You voted 57 to 1 on which post you preferred. Lines of code vetoed you. THE DATA 7,700 to 136 is 57 to 1. Adjust for impressions and it’s still 29 to 1 (See chart below) THE PRICE Not all engagement is equal. A like is free, and nobody sees you cast it. A repost staples my words to your name, in front of your followers. It’s the most expensive vote on this platform, and 7,700 people paid it. X wasn’t listening. THE MACHINE The ranker is Phoenix, a transformer built from Grok. It reads your post and predicts how likely people are to like, reply, or repost. Distribution flows from that prediction, made before a single human sees the post. A machine guesses what we want. Then it decides what we get. That machine learned our taste by watching us. It’s a copy of us. Now the copy outvotes the original, and when it gets us wrong, nothing tells it. That loop isn’t broken. It was never built. This is the design, and a design is a choice. Respectfully, I think the design should be revisited. THE OBJECTION Maybe the first post just found its ceiling. Maybe 198,000 views was all the audience it had. Except 7,700 reposts isn’t a ceiling, it’s a battering ram. Those views weren’t Phoenix being generous. They were 7,700 people carrying the post by hand. I watched it climb minute by minute. No algo push. A crowd. THE SUGGESTION Monday worked because someone inside noticed the gap. Engineers see the code. Users see what it does to us. Each side holds half the picture. This post is my half. Elon already said it: “We need a complete overhaul of the algorithm.” Start here. Stop letting a robot predict what we’ll like. Use what we already told you we like. One is a guess. The other is actual humans. PS: Supporting data is in the comments. Please don’t throttle my account. I’m an active user and content creator, and I’m trying to be helpful.

Graham Platner says the quiet part out loud in “previously unseen footage.” “Me dropping out is not an admission of guilt, it an admission that I got caught.” 👇🏻