X ALGORITHM NEWS đ¨
X just added a second way to score posts to its open source algorithm. It's switched off (for now). If X flips it on, posts people actually stay on would pull way ahead (so small accounts: write to be read, not skimmed)
tl;dr:
- today a post's score is a sum: likes + replies + time spent + the rest
- the new formula multiplies: everything else à time spent à how unlikely people are to scroll past
- in X's own test, time spent alone tripled the score
- the setting is still on the old way, nothing changes today
Full details:
- How it works today
X guesses how likely each person is to like, reply, share, stay and so on. Each guess gets a number and X adds them all up.
Time spent is one item on that list. A small one.
(I check the open source algo daily with Claude and share updates with you)
- How the new one would work
Three parts, multiplied together:
1. everything people might do with your post: likes, replies, shares, follows
2. a time part that grows the longer X thinks people will stay
3. a skip part that grows the less likely people are to scroll past
Multiplying is the big deal. A like on a post people stay on would count for more than a like on a post they skim.
- What X's own test shows
X wrote a test for the new formula. In it, a post people are expected to stay on for 20 seconds gets its score tripled, from time alone.
Those are test numbers. X hasn't said what it would use for real.
- Is it on?
No. The setting still says the old way, and the skip part is set to do nothing.
- My read
Every change since October points the same way. Likes up, time spent up, taps down, copy link gone. This is the same idea taken a lot further.
- Why it would matter if you're small
You can't win on reach yet. You can win on the people who do see your post staying on it. Under this formula that's worth more than it's ever been.
If it goes on, you'll hear it from me that day. Until then, write posts people don't want to leave.