One Article, Two X Accounts, Two Different Jobs

One Article, Two X Accounts, Two Different Jobs

user avatar

by me1afon

17 minutes ago


AI tools helped gather research and write the first draft of this article. I edited it and checked every figure and quote against the page it comes from, listed at the end. The cover illustration is AI-generated. Details at the end.

In April 2026, Nieman Lab compared two posts on X about the same investigation into the identity of Bitcoin's creator. One came from John Carreyrou, a reporter. The other came from The New York Times, an account with 53 million followers. The reporter's post, published first, had been liked, replied to or reposted 9,242 times; the newspaper's post, published a few hours later, 8,361 times [6].

That is one pair of posts, not a study, and the Nieman Lab author notes that both were "just a couple of sentences and a link" [6]. But it shows something worth taking seriously. A person and an institution can carry the same story to the same platform, and an audience of 53 million does not by itself settle which post readers respond to.

When one article goes out through a personal account and a publication's account, the two posts should do two different jobs. The personal account offers a person's argument and invites a reply. The publication's account offers a reference that is easy to save, recognise and come back to. Posting the same text twice wastes one of the two, and X's own developer rules say not to do it.

This is the second part of Platforms & Metrics. The first part covered adapting an article for LinkedIn. This one stays on X and on one question: how to split the work between an author's own account and the account of the publication.

What the feed code says about any account

X publishes the code that assembles its For You feed in a public repository, xai-org/x-algorithm. The README describes the core of the ranking simply: a model predicts the probability of about 25 different actions, from likes and replies to blocks and reports, and the final score is the sum of each weight times the predicted probability of that action [1]. The repository also warns against reading the weights as counts: they multiply a viewer's predicted probability, so it would be "incorrect" to conclude that one report "cancels out" a fixed number of likes [1].

In the version read for this article (commit 30647a8 of 14 November 2026), the default weights for positive actions include [2]:

  • reply 5.0, quote 5.0, share by direct message 5.0;
  • follow the author 4.0;
  • share 2.118, repost 1.0, like 0.708;
  • open a link 0.2, profile click 0.

These are defaults, and they multiply probabilities from a model whose predictions for any given post are not visible in the code, so they show what the system is built to value, not a formula for one post. They also change. The parameters file was updated in three commits between 12 and 14 November 2026 [2]. In the version this site's research read on 4 November, copying a post's link had the largest positive weight; in the version read for this article, that weight is zero [2][7]. Any advice built on a single weight should carry the date it was read.

One thing the ranking parameters and the README read for this article do not contain is a separate rule for company accounts [1][2]. As far as these files show, a post from @Dappexpert and a post from a person are scored by the same machinery. That does not prove nothing else in X's systems treats them differently. It does mean the difference between the two accounts has to come from somewhere else.

Where a person and a publication really differ

The difference comes from the follow graph: who follows whom, and in which direction.

Mutual follows. In July 2026 X added a boost for original posts from people a viewer follows and who follow the viewer back. The repository's own explanation says the boost raises "the weight on the predicted probability that you'll reply" to such a post [4]. It was tested at values from 5 to 20, rolled out at 20 to many users on 13 July and set to 15 on 24 July [4]; the current default is 15, added to the ordinary reply weight of 5 [2]. A person who follows and talks with others in a field builds exactly this kind of two-way connection. A publication's account usually follows few of its readers back, so for most of its audience this boost does not apply. That last point is an inference about typical accounts, not something the code says.

Reposts and replies do not travel to strangers. By default, the feed drops reposts and replies from accounts the viewer does not follow [3]. The README describes the same filter: "Reposts and replies from accounts the viewer does not follow" are removed [1]. So if the publication's account reposts the author's post, the repost reaches the publication's own followers and nobody else through the For You feed. A reply under someone else's large post is seen inside that conversation, not pushed to new readers on its own. To reach people who follow neither account, each account needs an original post of its own.

Bursts work against the author. The README lists an author diversity adjustment: "each post after an author's first is multiplied by a decaying factor, down to a floor" [1]. Splitting one article into many quick posts from one account mostly competes with itself.

None of these rules is about personal versus company accounts. Together they explain why the two accounts end up doing different work. The personal account is where two-way conversation happens and where replies are most valuable. The publication's account speaks mostly to its own followers, so its job is to make each post worth keeping and to make the next one predictable enough to follow.

 

The personal account: an argument with a person behind it

On Dapp.Expert, the personal account is @mnemath1, the author's own. Its post for an article is an adapted text with the verified link, and a short video when one is ready; until then, the approved cover image, the same one used on LinkedIn.

The job of this post is to start a conversation that a knowledgeable reader can join. That suggests a shape:

  1. One claim, in the author's words. Not the headline, but the point the author would defend in a reply.
  2. The reason, briefly. One number or one finding, with where it comes from.
  3. An honest limit. Where the claim could be wrong. This gives a reader who knows more a reason to answer rather than scroll on.
  4. The link, after a plain line. The line is "Read it on Dapp.Expert:", with the address on the next line. On X, an invisible zero-width space is typed after "Dapp", so that X does not turn the name into a second link and the post carries a single link.

A reply and a follow are each weighted several times higher than a like [2]. The more practical reason is that this is the account a reader can actually talk to. A press-release tone throws away the one thing it has and the other account lacks.

The publication's account: a reference worth keeping

The publication's account is @Dappexpert. Its post for an article carries a short text with the link and four numbered images, 1/4 to 4/4, each card explaining one point from the article: the question, the evidence, the data, and what to do with it.

The job here is different. A reader who follows a publication expects a recognisable product: the same series label, the same layout, sources named on the card, and the same quality next time. A follow, weighted at 4.0 [2], is in effect a bet that the next post will be useful in the same way. The account earns it by being consistent rather than by being personal.

A practical lesson from preparing these cards: the X composer and the published post did not crop a multi-image post the same way. In one earlier post, the composer showed a neat grid, while the live post showed a horizontally scrolling carousel that cut the edges of wide cards. The cards for this series are now square, with the text kept in a central safe area, and the live post is checked separately from the preview. As with LinkedIn in the first part, the rule is to read the published post, not the draft.

Why the two posts must not be copies

There is also a plain rule. X's Developer Policy, which applies to services that post through X's interfaces, says: "Never post identical or substantially similar content across multiple accounts" [5]. The same section of the policy points to X's Automation Rules and its rules on platform manipulation and spam [5]. X's help-centre pages on automation could not be opened for this article, because they block automated access, so their exact wording is not quoted here.

For a small publication that schedules posts with tools, the safe reading is simple: the two accounts can share the article and the link, but not the text. A different angle, wording and media keep the posts from being duplicates, and they are also what let each one do its own job.

A worked example

The previous part of this series, on turning an article into a LinkedIn post, was prepared with two X posts. Here is how they divide the work.

The personal post opens with a claim in the author's voice: "The LinkedIn 'link penalty' is mostly a pointer penalty." It then gives two numbers from the vendor dataset the article relies on: a 35% gap when the link preview card was the only visual, and 2% when the address sat in a text post. It ends with one sentence of advice and the link. A reader can disagree with the claim, and that is the point.

The publication's post opens with a description of what the reader gets: "How to turn an article into a LinkedIn post that works without the click. Four cards: what LinkedIn measures, what the link data shows, and a five-part structure." The four cards then carry the reference material: the question, LinkedIn's own description of its signals, the dataset's numbers with the sample size, and the five-part structure.

The two posts share a link and a subject. They share no sentences. One is something to answer; the other is something to save.

Links, briefly

Both accounts carry the link to the original article, because the original is the point. There is evidence that a publication account made almost entirely of links has a harder time on X. In the same Nieman Lab analysis, which looked at the 200 most recent posts from each of 18 accounts, links appeared in 88% of The New York Times' posts and 98% of The Wall Street Journal's, while Fox News linked in only 9% and had the third-highest median engagement in the sample [6]. The author is careful to add that links "are not the only thing" holding big publishers back [6]; account size, subject and style differ too. The fourth part of this series looks at external links and reach directly.

What to measure for each account

Because the jobs differ, the measures should too.

  • Personal account: replies and new follows relative to impressions. If those rise, the posts are starting conversations. This site's research notes suggest new follows per 1,000 impressions as a simple ratio to watch [7].
  • Publication account: saves, shares and follows, and whether readers who arrive from a card post stay on the article.
  • Both: visits to the article itself. Dapp.Expert's author posts use a tagged link: utm_source=x, utm_medium=social, utm_campaign set to the article's identifier, and utm_content naming the version. Giving each account its own utm_content value shows which post sent readers who actually read.

Later parts of this series look at why view counts from different accounts and platforms cannot simply be added together.

A short check before you post

  • Does each account's post have its own text, not a copy of the other?
  • Does the personal post make a claim the author would defend in a reply, with one honest limit?
  • Does the publication's post look like the rest of its series, with sources visible?
  • Is the link in each post after a plain line?
  • Is each account posting an original post, rather than one reposting the other?
  • Did you check the live post, not only the composer preview?

 

What this article cannot tell you

The weights come from public code at one point in time; they multiply model predictions the code does not show, can be overridden in experiments and have changed within days. Nothing here shows that X treats company accounts and personal accounts identically everywhere, only that the files read for this article contain no separate rule for them. The point that publication accounts rarely benefit from the mutual-follow boost is an inference about typical accounts. The Carreyrou and New York Times pair is a single example, and the Nieman Lab link figures describe 18 accounts and show a pattern, not a cause. X's help-centre rules on automation could not be read directly. Nothing here reports measured results from Dapp.Expert's own X accounts.

How this article was made

Task What the AI did What a person did How it was checked
Research Collected X's open-source code analysis and platform studies into research notes Chose the question and the sources to rely on Weights and filters re-read in the repository files at commit 30647a8 [1]–[4]; policy and study quotes checked on their pages [5], [6]
Draft Wrote the first draft from the notes Edited the text, structure and worked example A separate review compared figures and claims with the sources
Images Generated the photograph-style illustration on the cover Chose the illustration; designed the explanatory diagrams Labelled as AI-generated in the disclosure line and alt text
Not done by AI — The decision to publish and responsibility for the text —

Conflicts of interest: none. The author has no commercial relationship with X or with the organisations whose work is cited.

Responsible for this article: Dmitriy Lastovka, author. Corrections: contact form.

Sources

Labels: code (X's published source code, read in the repository), official (the platform's own page), study (a published analysis, read on its own page).

  1. xai-org/x-algorithm, README, commit 30647a8 (14 November 2026). https://github.com/xai-org/x-algorithm Code.
  2. xai-org/x-algorithm, home-mixer/params/param.rs, commit 30647a8 (14 November 2026), and its commit history. https://github.com/xai-org/x-algorithm/blob/main/home-mixer/params/param.rs Code.
  3. xai-org/x-algorithm, home-mixer/filters/oon_retweet_reply_filter.rs. https://github.com/xai-org/x-algorithm/blob/main/home-mixer/filters/oon_retweet_reply_filter.rs Code.
  4. xai-org/x-algorithm, "Example algorithm diff: bidirectional follow boost", docs/BIDIRECTIONAL_BOOST_CHANGE.md. https://github.com/xai-org/x-algorithm/blob/main/docs/BIDIRECTIONAL_BOOST_CHANGE.md Code and official.
  5. X Developer Policy. https://docs.x.com/developer-terms/policy Official.
  6. Laura Hazard Owen, "Do links hurt news publishers on Twitter? Our analysis suggests yes", Nieman Lab (8 April 2026). https://www.niemanlab.org/2026/04/do-links-hurt-news-publishers-on-twitter-our-analysis-suggests-yes/ Study.
  7. Dapp.Expert research notes on X's feed code (4 November 2026) and on platform adapters (5 November 2026). Author's research; not a published source.

Tier I

Sector: #18291

Sealed Hiding Place Room

Resource Cache

Resource Cache

Tier I

Requires 25% Tier Progress to Claim
Meme Cache

Meme Cache

Tier I

Requires 50% Tier Progress to Claim
Equipment Cache

Equipment Cache

Tier I

Requires 75% Tier Progress to Claim

After collecting, hiding places will be stored in your inventory and can be opened with Keys.

Other articles

Turning an Article Into a LinkedIn Post

A LinkedIn post that only points to an article gives readers nothing to stay for. What LinkedIn measures, what link data shows, and a post that stands alone.

user avatarme1afon

Comparing Creators Without Rewarding Only Incumbents

A ranking by totals mostly measures who started first. Why early leads persist, and five corrections to compare creators fairly without punishing leaders.

user avatarme1afon

Making an Expert Ranking Explainable

A ranking of experts is a claim about real people. Four parts make it checkable: a published formula, stated weights, a per-entry breakdown and an appeal path.

user avatarme1afon

Audience Size Is Not Topic Authority

A large following records popularity, not knowledge of a subject. Why reach and authority need separate evidence, and five questions to ask about any expert.

user avatarme1afon

Spotting Gaming in Creator Program Metrics

Any number that pays creators attracts people who produce the number, not the work. Three ways metrics get gamed, seven signs, and why a sign is not proof.

user avatarme1afon

How Creator Incentives Change What Gets Published

Six documented cases show how platform incentives change what creators publish, plus five questions to ask before you design or join a creator programme.

user avatarme1afon

Important disclaimer: The information presented on the Dapp.Expert portal is intended solely for informational purposes and does not constitute an investment recommendation or a guide to action in the field of cryptocurrencies. The Dapp.Expert team is not responsible for any potential losses or missed profits associated with the use of materials published on the site. Before making investment decisions in cryptocurrencies, we recommend consulting a qualified financial advisor.