For posts and threads on X

Worth a ReplyX published what its ranker counts. Write the post it counts.

In 2023 X open-sourced the weights its timeline ranker gives each engagement: a reply weighs 13.5, a reply the author answers 75, a like 0.5, a report -369. Paste a draft: the page counts it the way X does and shows what it is set up to earn, free. One run says which engagements it invites or risks and writes three rewrites, each aimed at a different one.

Try a sample:

1 Your draft free

The post - or a thread, split by a line of --- or numbered 1/ 2/ 3/

0 / 280

What you want from it Who will see it

2 As it would read free

3 What the published ranker counts free

Each candidate post gets a score: the sum of each weight times the predicted chance of that engagement. After a run, each row shows whether your draft invites it, risks it, or leaves it alone.

4 The tune metered

One run reads the draft against all ten engagements, gives a verdict, writes three rewrites of the first post - each aimed at a different engagement - and says what to do once replies come in.

Pricing as soon as the session is ready...

The reply could not be read. Here it is as it came back:

NOT TUNED YET

Three rewrites

Once replies come in

The page's check

Nothing is posted from this page. "Post on X" opens X's own compose box with the rewrite; make it sound like you and post it there.

5 Saved posts

How it works

The count follows X's open-source counting rules: most Latin-script characters count 1, Chinese, Japanese, Korean and other wide scripts count 2, every emoji (with its skin tone or joined parts) counts 2, and every link counts 23 however long it is. A post over 280 needs a paid X account to send.

The weights come from the README of the home-timeline heavy ranker X open-sourced in 2023 (twitter/the-algorithm-ml, projects/home/recap). That README says the weights were first set so each engagement adds about the same to the score on average, and that they are adjusted over time - so they show what the model counts, not what one reply is worth today, and no page can predict your reach.

The run's reply is checked: its verdict must follow from its own pulls (a risk of mutes or reports makes it risky; no reason to reply or click in makes it a rework), a cost can never be "invited", and every rewrite is re-counted and refused if it adds a number, a mention, a link or a hashtag your draft never had, stacks hashtags or asks for engagement outright.