Sabitlenmiş Tweet
The new X algorithm doesn’t reward the time you post, it rewards how your post is scored.
For years, creators were told:
“Post at 4am to 9am.”
“Use threads.”
“Engage more".
Those works no doubt but the way X distributes content has evolved far beyond those tips.
When parts of X’s recommendation system were open-sourced, something important became clear: The logic behind distribution is technical, measurable and predictable if you understand what the system is optimizing for.
@workagnt studied that system, then turned it into an AI employee.
What actually happens after you hit “Post”?
Your content doesn’t simply go to followers.
It enters a ranking pipeline known internally as Home Mixer, which evaluates where and to whom your post should appear.
Two systems handle this:
Thunder → prioritizes delivery to people who already follows you (in-network)
Phoenix → a machine-learning ranking engine that decides if non-followers should see your post (out-of-network)
And here’s the key insight:
Your post is scored on its own merits, not against other posts or based on time of day but on how likely it is to generate engagement.
It either earns distribution or it doesn’t.
What the ranking model is predicting, the system is trained to estimate engagement probability.
In order of impact, the signals look like this:
➩ Replies (conversation)
➩ Reposts (share value)
➩ Likes (low-friction approval)
➩ Bookmarks (utility)
➩ Dwell time (attention)
This is why smaller accounts can outperform larger ones.
The model doesn’t rank people based on follower count, it ranks posts.
What quietly reduces your reach?
Based on how quality filters and ranking rules are structured:
➩ Posts that push external links are deprioritized
➩ Hashtag overuse is flagged as spam behavior
➩ Engagement bait is detectable and penalized
➩ Posting too frequently hurts individual post performance
➩Commonly muted words reduce distribution
➩Negative feedback (mute, block, not interested) trains the system against similar posts
Many creators trigger these filters without realizing it.
Why threads and media consistently perform well?
➩ Each reply in a thread becomes a new ranking candidate
➩ Images and videos supply additional signals into the model
➩ Prior interactions and mutual follows affect in-network priority
➩ Community Notes can suppress reach if applied
Threads work because they create multiple scoring opportunities.
Inside @workagnt, they built an AI employee trained on how X ranks content.
You paste your draft post.
It evaluates:
➩ Hook strength (for dwell time)
➩ Likelihood of generating replies vs passive likes
➩ Hidden reach killers (links, hashtags, bait patterns)
➩ Whether the post should be a thread
➩ Specific rewrites explained through ranking logic
So before you publish, you understand how the algorithm is likely to interpret your post.
Most people are still posting based on outdated advice.
You can now post based on how X actually works today.
English
