<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/">
  <channel>
    <title>Twitter on Liam Ding</title>
    <link>https://www.liamding.com/en/tags/twitter/</link>
    <description>Recent content in Twitter on Liam Ding</description>
    <image>
      <title>Liam Ding</title>
      <url>https://www.liamding.com/img/og-cover.png</url>
      <link>https://www.liamding.com/img/og-cover.png</link>
    </image>
    <generator>Hugo -- 0.147.7</generator>
    <language>en</language>
    <lastBuildDate>Tue, 17 Nov 2026 00:00:00 +0800</lastBuildDate>
    <atom:link href="https://www.liamding.com/en/tags/twitter/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>X&#39;s clip button ships the credit, not the consent</title>
      <link>https://www.liamding.com/en/posts/2026-11-17-x-clipping-ships-credit-not-consent/</link>
      <pubDate>Tue, 17 Nov 2026 00:00:00 +0800</pubDate>
      <guid>https://www.liamding.com/en/posts/2026-11-17-x-clipping-ships-credit-not-consent/</guid>
      <description>X&amp;#39;s native clipping attaches credit and a timestamped link to every clip, and each side earns only if that side is enrolled. But creators can&amp;#39;t turn it off yet, and the only control they have is deleting the original.</description>
      <content:encoded><![CDATA[<h2 id="the-problem">The problem</h2>
<p>On 9 November X shipped native clipping. Tap Repost, then Clip video, pick 5 to 60 seconds of someone else&rsquo;s video, add your take and post it (<a href="https://x.com/XCreators/status/2119892172757491898">@XCreators</a>). Every clip carries a &ldquo;Clip from&rdquo; label that opens the original at the moment the clip starts, and both the clipper and the original creator can earn from the clip through Original Content Rewards (<a href="https://x.com/XCreators/status/2119892170224074832">@XCreators</a>).</p>
<p>Allegra Jacchia, X&rsquo;s Head of Creator Product (<a href="https://www.netinfluencer.com/x-launches-native-clipping-paying-both-clippers-and-original-creators/">NetInfluencer</a>), framed it as a fix for how clipping used to work on X: download the video, edit it in another tool, re-upload it, and the original creator &ldquo;often didn&rsquo;t get credit, and even when a clip went viral, they didn&rsquo;t earn from it&rdquo; (<a href="https://x.com/allegrajacchia/status/2119893586720501974">Allegra Jacchia</a>).</p>
<p>Most coverage stops at &ldquo;both sides get paid&rdquo;. The more useful question for anyone posting long video on X is what the rules decide on the creator&rsquo;s behalf: who can cut your video, how much of it, what you get back, and what you can do if you don&rsquo;t like it.</p>
<h2 id="what-i-concluded">What I concluded</h2>
<p>X built the credit into the clip and left the consent out. Attribution is attached automatically: the help page says &ldquo;The original creator is credited on every clip,&rdquo; and every clip has a &ldquo;Clip from&rdquo; label plus a link back to that moment in the original. Permission is not part of the design yet. The help page&rsquo;s sentence is: &ldquo;You cannot turn off clipping for your videos right now, but we are working on adding that option. If you delete your original video, any clips of it are removed too.&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>)</p>
<p>The money is narrower than the headline. Each side earns only if that side is enrolled in Original Content Rewards. For the clipper, &ldquo;engagement on your clip counts toward your Original Content Rewards earnings, the same as your other posts&rdquo;; for the original creator, &ldquo;These earnings are calculated the same way as earnings from embedded reposts of your content&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>). Neither the clipping help page nor the Original Content Rewards help page states a rate for a clip or a split between the two sides. ContentGrip notes the same gap, and is only corroboration here (<a href="https://www.contentgrip.com/x-video-clipping-payouts/">ContentGrip</a>). For a creator who is not enrolled, clipping is a credit-and-traffic feature, not an income feature.</p>
<p>So the real protection for the creator is a length rule, not a permission: a clip can be at most 30% of the original. That cap, plus the link back, is X&rsquo;s bet that a clip will send viewers to the full video rather than replace it.</p>
<h2 id="constraints">Constraints</h2>
<ul>
<li><strong>Which videos.</strong> Pre-recorded videos that are at least 2 minutes long; live broadcasts can&rsquo;t be clipped yet (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>). The launch thread words it as &ldquo;any video longer than 2 minutes&rdquo; (<a href="https://x.com/XCreators/status/2119892172757491898">@XCreators</a>).</li>
<li><strong>How long a clip.</strong> 5 to 60 seconds and no more than 30% of the original. X&rsquo;s own examples: a clip from a 2-minute video can be up to 36 seconds, and videos longer than 3 minutes 20 seconds can be clipped for the full 60 (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>). That 3:20 is just 60 ÷ 0.3 = 200 seconds.</li>
<li><strong>Where.</strong> The iOS and Android apps, with &ldquo;Web is coming soon&rdquo; and no date; clipping &ldquo;may vary by app version and region&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>).</li>
<li><strong>Who earns.</strong> Each side earns only if that side is enrolled. The help page says: &ldquo;To earn from clips, you need to be enrolled in Original Content Rewards.&rdquo; For the original creator: &ldquo;These earnings are calculated the same way as earnings from embedded reposts of your content.&rdquo; On payouts: &ldquo;clip earnings are included in your regular Original Content Rewards payouts, starting with the first payout after clipping launches.&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>) Joining also requires that you &ldquo;Be based in a country where the Original Content Rewards Program is available.&rdquo; The page says: &ldquo;Meeting these requirements does not guarantee admission to the Program. Once you meet all eligibility requirements, you can submit a quick application to be considered for the Program.&rdquo; The listed bar includes &ldquo;Hold an active X Premium, Premium+, or Premium Business subscription, or a Plus, Premium, Super, or Ultra plan on The X Subscription,&rdquo; &ldquo;Have at least 500 verified followers,&rdquo; and &ldquo;Have at least 500,000 Home Timeline impressions from verified users in the last 90 days (impressions on replies are excluded).&rdquo; The same list also requires &ldquo;Be 18+ years of age,&rdquo; &ldquo;Have an X account in good standing with no history of repeatedly violating our Monetization Guidelines or Terms of Service,&rdquo; &ldquo;Have a Personal or Business Account&rdquo; (&ldquo;Accounts belonging to political or government organizations are not eligible.&rdquo;), and &ldquo;Actively post original content as defined in this policy.&rdquo; (<a href="https://help.x.com/en/using-x/original-content-rewards">Original Content Rewards</a>).</li>
<li><strong>What it doesn&rsquo;t count for.</strong> &ldquo;Clips do not count toward the Video Kickstart Bonus, because a clip is a post of someone else&rsquo;s video.&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>) That is the invite-only posting bonus I wrote about <a href="/en/posts/2026-11-10-x-video-kickstart-bonus-pays-for-showing-up/">last week</a>.</li>
<li><strong>Native clip vs re-upload.</strong> Original Content Rewards counts a download-and-reupload as Copied content: &ldquo;This includes content that is downloaded from X or other platforms and then reuploaded onto X.&rdquo; The same page says that content does not pay out: &ldquo;It is downloaded from X or another platform and reuploaded onto X (unless you are the author of that content)&rdquo; (<a href="https://help.x.com/en/using-x/original-content-rewards">Original Content Rewards</a>). Clip video is a different kind of post on the clipping help page, not the download-and-reupload this sentence describes.</li>
</ul>
<h2 id="the-options-and-what-i-traded-away">The options, and what I traded away</h2>
<p>For a creator posting long video on X today:</p>
<p>The options below are the ones I listed from the rules X has published. They are not X&rsquo;s recommendations.</p>
<table>
  <thead>
      <tr>
          <th>What you do</th>
          <th>What you get</th>
          <th>What it costs</th>
      </tr>
  </thead>
  <tbody>
      <tr>
          <td>Leave clipping on (the default, and for now the only setting)</td>
          <td>Credited clips that link back to the exact moment; a share of clip earnings if you are enrolled</td>
          <td>No say over who clips you or what commentary sits on top of your 60 seconds</td>
      </tr>
      <tr>
          <td>Keep videos under two minutes</td>
          <td>Your videos can&rsquo;t be clipped natively</td>
          <td>A format constraint you didn&rsquo;t otherwise want; nothing stops a download-and-re-upload</td>
      </tr>
      <tr>
          <td>Delete a video after a bad clip</td>
          <td>Every clip of it disappears</td>
          <td>The original&rsquo;s views, replies and links go with it, and so does every good clip</td>
      </tr>
      <tr>
          <td>Post the long version live</td>
          <td>Livestreams can&rsquo;t be clipped yet</td>
          <td>Only until X extends clipping to live, which a user asked for and Jacchia answered &ldquo;It&rsquo;s coming :)&rdquo; (<a href="https://x.com/allegrajacchia/status/2119920838141166044">Allegra Jacchia</a>)</td>
      </tr>
  </tbody>
</table>
<p>I would take the first row and design for it rather than fight it. The other three rows trade away more than a clip costs, and the third is a nuclear option, not a control.</p>
<h2 id="evidence">Evidence</h2>
<p><strong>What X says on its own accounts.</strong> The @XCreators launch post says every clip &ldquo;credits your original video and links back to it&rdquo; and that &ldquo;both the original creator and the clipper can earn from views on the clip through Original Content Rewards&rdquo; (<a href="https://x.com/XCreators/status/2119892170224074832">@XCreators</a>). The follow-up gives the flow and limits: Repost, then Clip video; iOS and Android; videos longer than 2 minutes; clips of 5 to 60 seconds (<a href="https://x.com/XCreators/status/2119892172757491898">@XCreators</a>). Jacchia&rsquo;s post adds the before-and-after and says that &ldquo;for the first time&rdquo; both sides can earn (<a href="https://x.com/allegrajacchia/status/2119893586720501974">Allegra Jacchia</a>).</p>
<p><strong>What the help page says.</strong> X&rsquo;s help page, &ldquo;About clipping on X&rdquo;, opens with: the original creator &ldquo;is credited on every clip, and creators enrolled in Original Content Rewards earn from it&rdquo; (<a href="https://help.x.com/en/using-x/clipping">X Help Center</a>). I read the rules and the FAQ directly in a browser on 16 November 2026. The answer to &ldquo;Can I stop people from clipping my videos?&rdquo; is &ldquo;Not right now. We are working on an option to turn off clipping for your videos.&rdquo; The answer to &ldquo;What happens to clips if I delete my video?&rdquo; is &ldquo;When you delete the original video, any clips of it are removed too.&rdquo; The answer to &ldquo;Do clips count toward the Video Kickstart Bonus?&rdquo; is &ldquo;No. Clips are posts of someone else&rsquo;s video, so they do not count as qualifying videos.&rdquo; Under &ldquo;Earning from clips&rdquo;, the original creator&rsquo;s earnings &ldquo;are calculated the same way as earnings from embedded reposts of your content.&rdquo; The answer to &ldquo;Do I earn from clips I make?&rdquo; is &ldquo;Yes, if you are enrolled in Original Content Rewards. Clips earn like your other posts, and the original creator earns too if they are enrolled.&rdquo; NetInfluencer and ContentGrip summarise the same page in line with these sentences, and are only corroboration here (<a href="https://www.netinfluencer.com/x-launches-native-clipping-paying-both-clippers-and-original-creators/">NetInfluencer</a>, <a href="https://www.contentgrip.com/x-video-clipping-payouts/">ContentGrip</a>).</p>
<p><strong>What nobody has published.</strong> Neither the clipping help page nor the Original Content Rewards help page states a rate per clip view or a split between clipper and creator. The clipping page only points the original creator&rsquo;s formula at &ldquo;earnings from embedded reposts&rdquo;. The rewards page says &ldquo;Eligible creators earn payouts based on qualified impressions generated by their original content&rdquo; and &ldquo;Payouts are currently processed every two weeks and the minimum payout is $30&rdquo; (<a href="https://help.x.com/en/using-x/original-content-rewards">Original Content Rewards</a>). ContentGrip notes that X has not published a per-view rate or a split, as corroboration (<a href="https://www.contentgrip.com/x-video-clipping-payouts/">ContentGrip</a>). Any per-clip income figure you see on X right now is someone&rsquo;s estimate, not X&rsquo;s number.</p>
<p><strong>The design choices, read together.</strong> Credit is attached automatically and travels with the object. Payment reuses an existing formula (embedded reposts) instead of a new one. The protection is quantitative (30%, 60 seconds) rather than a permission. And the one control the creator does have, deletion, is all-or-nothing. These choices look like they keep the feature easier to ship. Together, it may be that the clipper decides which moment of your video represents you, and you may only find out after it is posted.</p>
<h2 id="where-this-stops-being-true">Where this stops being true</h2>
<ul>
<li><strong>When the opt-out ships.</strong> X says it is working on one. If it is per-video rather than per-account, the trade-off table above changes most.</li>
<li><strong>If X publishes a rate or a split.</strong> &ldquo;calculated the same way as earnings from embedded reposts of your content&rdquo; is a pointer to an existing formula, not a number. A published figure would let creators compare clip income with other formats.</li>
<li><strong>When clipping reaches the web and livestreams.</strong> Today&rsquo;s workaround of going live instead of uploading disappears.</li>
<li><strong>The help page can change.</strong> The length limits, the opt-out, deletion and payout details above are from &ldquo;About clipping on X&rdquo; and the Original Content Rewards help page as read on 16 November 2026. When those pages change, these details go stale.</li>
</ul>
<h2 id="what-i-would-do-differently">What I would do differently</h2>
<p>If I were posting long video on X now, I would assume every video over two minutes will be clipped and edit for that. Put the strongest 30 to 60 seconds after a clear marker so the &ldquo;Clip from&rdquo; timestamp lands somewhere that makes sense out of context. Make videos at least 3:20 long if you want clippers to be able to take a full minute. Put the context a cold viewer needs in the video, not only in the post text, because the clipper&rsquo;s commentary replaces yours. If you are in a country where Original Content Rewards is available and you are near the other requirements, submit the application before clips of you start circulating, since only enrolled creators earn from them.</p>
<p>If I were clipping, I would use the Clip video button and never download and re-upload: the native clip keeps the credit visible, can earn, and avoids the help page&rsquo;s Copied content, &ldquo;content that is downloaded from X or other platforms and then reuploaded onto X.&rdquo; And I would not budget around clip income until X publishes a number.</p>
]]></content:encoded>
    </item>
    <item>
      <title>X&#39;s Video Kickstart Bonus pays for showing up, not for a hit</title>
      <link>https://www.liamding.com/en/posts/2026-11-10-x-video-kickstart-bonus-pays-for-showing-up/</link>
      <pubDate>Tue, 10 Nov 2026 00:00:00 +0800</pubDate>
      <guid>https://www.liamding.com/en/posts/2026-11-10-x-video-kickstart-bonus-pays-for-showing-up/</guid>
      <description>X&amp;#39;s invite-only Video Kickstart Bonus counts 15 posting days, not views. Read as a spec it is a payment for attendance with a final review that can still strike any video, and it says which behaviour X wants more of.</description>
      <content:encoded><![CDATA[<h2 id="the-problem">The problem</h2>
<p>X&rsquo;s Video Kickstart Bonus is an invite-only pilot: post 15 qualifying videos on 15 different days in a program period and you earn a bonus for that period (<a href="https://web.archive.org/web/20261015220634/https://help.x.com/en/using-x/video-kickstart-bonus">X Help Center</a>). Trade press has framed it as X following YouTube into paying for consistency (<a href="https://www.tubefilter.com/2026/11/03/x-video-kickstart-program-creator-bonus-monetization/">Tubefilter</a>), and the first period ends on 14 November. Most coverage repeats the headline rule and a dollar figure.</p>
<p>The more useful question is what the rules actually pay for. Most creators will never get an invite, so the spec matters mainly as a statement of what X&rsquo;s creator team wants more of, and as a test of whether a cadence like this is worth copying on your own account.</p>
<h2 id="what-i-concluded">What I concluded</h2>
<p>The bonus pays for attendance. The help page lists no view, like or follower threshold for the period, and the amount is not a function of performance: it depends on your total followers across X and other platforms at the time you are invited, in a range that stays fixed for the whole program (<a href="https://web.archive.org/web/20261015220634/https://help.x.com/en/using-x/video-kickstart-bonus">X Help Center</a>). The Terms add that extra videos above 15 do not increase the reward (<a href="https://web.archive.org/web/20261016072644/https://legal.x.com/en/x-creator-kickstart-bonus-terms">Terms</a>).</p>
<p>That makes it different in kind from Original Content Rewards, which pays on qualified impressions and asks, at application, for at least 500,000 Home Timeline impressions from verified users in the last 90 days and 500 verified followers (<a href="https://web.archive.org/web/20261019152944/https://help.x.com/en/using-x/original-content-rewards">Original Content Rewards</a>). Kickstart reads as aimed at creators who may not have cleared that bar yet: it pays for a posting habit, and fast-tracks the creator into the performance program.</p>
<p>&ldquo;Pays for attendance&rdquo; is only half the picture, because the Terms keep a lot of discretion on the other side. Showing up makes you eligible, and X&rsquo;s reviewers decide the rest.</p>
<h2 id="constraints">Constraints</h2>
<ul>
<li><strong>Both goals in one period.</strong> You need 15 qualifying videos and 15 different posting days. A posting day is a calendar day in Pacific Time, and several videos on one day still count as one day. X&rsquo;s own examples: 15 videos on one day fails; 30 videos across 10 days fails; 20 videos across 14 days fails (<a href="https://web.archive.org/web/20261015220634/https://help.x.com/en/using-x/video-kickstart-bonus">X Help Center</a>).</li>
<li><strong>The video itself.</strong> 9:16 vertical, under 3 minutes, yours, posted from the enrolled account after you enrol. Clips already posted on Instagram, TikTok or YouTube count, unless they carry another platform&rsquo;s watermark, name or logo.</li>
<li><strong>Only after enrolling.</strong> Period 1 runs from 15 October to 14 November 2026 (Pacific Time), with enrolment closing on 30 October (<a href="https://web.archive.org/web/20261016072644/https://legal.x.com/en/x-creator-kickstart-bonus-terms">Terms</a>). By my arithmetic, someone who enrolled on the last day had at most 16 calendar days left to land 15 posting days.</li>
<li><strong>The tracker is not the verdict.</strong> Creator Studio can lag by 48 hours, and its counts are estimates (<a href="https://web.archive.org/web/20261015220634/https://help.x.com/en/using-x/video-kickstart-bonus">X Help Center</a>). The Terms say the Final Count after review controls and can be higher or lower than any tracker.</li>
<li><strong>Payment is discretionary.</strong> The Terms call rewards discretionary, &ldquo;not a revenue share&rdquo;, and say meeting the criteria makes a creator eligible for consideration only. X may recoup payments, and may add, change, shorten or end periods.</li>
</ul>
<h2 id="the-options-and-what-i-traded-away">The options, and what I traded away</h2>
<table>
  <thead>
      <tr>
          <th>What you do with the spec</th>
          <th>What you get</th>
          <th>What it costs</th>
      </tr>
  </thead>
  <tbody>
      <tr>
          <td>Not invited: ignore it</td>
          <td>Nothing lost</td>
          <td>You miss a clear statement of which format X is pushing right now</td>
      </tr>
      <tr>
          <td>Not invited: copy the 15-days-in-a-month rhythm</td>
          <td>A cadence you can test against your own numbers</td>
          <td>Real production time; no bonus, no fast-track, no guarantee X&rsquo;s feed rewards it</td>
      </tr>
      <tr>
          <td>Invited: treat it as pay per day</td>
          <td>Predictable effort, an amount you can read in Creator Studio</td>
          <td>The review can still remove a video, so the effort is not fully yours to control</td>
      </tr>
      <tr>
          <td>Invited: minimum effort (15 clips, 15 days)</td>
          <td>Lowest cost to qualify</td>
          <td>Terms exclude spam, mass-produced posts made only to manufacture qualifying posts, and near-duplicates</td>
      </tr>
  </tbody>
</table>
<p>If I had an invite, I would take the third row and plan for the review, not the tracker.</p>
<h2 id="evidence">Evidence</h2>
<p><strong>What disqualifies a video, including after it showed as qualified.</strong> The help page lists: not your own content, a video downloaded from another X user and re-uploaded, a repost or quote post of someone else&rsquo;s video, another platform&rsquo;s watermark, name or logo, a claim by another rights holder, a duplicate of one you already posted, a Community Note rated Helpful, a paid partnership or third-party sponsored post, inauthentic engagement, and violations of the X Rules or Creator Monetization Standards. It also says a video that showed as qualified can still be disqualified at final review, for example through a rights-holder claim or a Helpful Community Note (<a href="https://web.archive.org/web/20261015220634/https://help.x.com/en/using-x/video-kickstart-bonus">X Help Center</a>). Two of those are decided by other people on their own schedule.</p>
<p><strong>Deleting costs a posting day.</strong> Deleted or removed videos, and videos hidden because the account became protected, stop counting. The Terms go further: qualifying posts must stay public through the end of the period and until bonus review is complete, and for a period in which they qualified, remain public and unedited for at least 30 days after the period ends (<a href="https://web.archive.org/web/20261016072644/https://legal.x.com/en/x-creator-kickstart-bonus-terms">Terms</a>). Deleting a flop is the natural instinct, and it is one of the moves the program penalises.</p>
<p><strong>What is excluded on purpose.</strong> Branded content, paid partnerships and promoted posts do not count, and the Terms rule out spam, posts mass-produced solely to manufacture qualifying posts, posts through unauthorised automation, and identical near-duplicates. A cadence program with an anti-cadence-hacking clause is telling you the 15 days are supposed to be 15 real videos.</p>
<p><strong>Reuse is allowed, borrowing is not.</strong> Your own clips from other platforms count if they are free of watermarks. The Terms say videos that are not yours do not count even if you have permission to reshare them as a clip or stitch (<a href="https://web.archive.org/web/20261016072644/https://legal.x.com/en/x-creator-kickstart-bonus-terms">Terms</a>). One press report says commentary built on third-party material can qualify (<a href="https://www.moneycontrol.com/technology/x-tests-video-kickstart-bonus-offers-creators-rs-2-85-lakh-for-15-videos-in-30-days-article-14055325.html">Moneycontrol</a>); I follow the Terms.</p>
<p><strong>The amount.</strong> X&rsquo;s help page does not publish a figure. The announcement post text does not contain one either (<a href="https://elonmuskarchive.org/posts/2110832477149577223">Allegra Jacchia&rsquo;s post</a>). NetInfluencer reported &ldquo;up to $3,000&rdquo; (<a href="https://www.netinfluencer.com/x-pilots-3000-usd-bonus-program-for-creators-who-post-vertical-video/">NetInfluencer</a>), Moneycontrol a fixed $3,000. Those two do not agree with each other, and neither is on X&rsquo;s own pages, so I treat the number as unconfirmed and the follower-range rule as the fact.</p>
<p><strong>What else the invite carries.</strong> A fast-track into Original Content Rewards and Premium+ free for one year if you do not have it, paid with the next Original Content Rewards payout (X Money in the US, a connected Stripe account elsewhere). The help page does not say what &ldquo;fast-tracked&rdquo; waives, so I do not assume the 500,000-impression gate is waived.</p>
<h2 id="where-this-stops-being-true">Where this stops being true</h2>
<ul>
<li>This is a pilot. X may add, change or end periods at any time, and has made no commitment to run another one (<a href="https://web.archive.org/web/20261016072644/https://legal.x.com/en/x-creator-kickstart-bonus-terms">Terms</a>). A later period could add a performance condition.</li>
<li>I read the help page as archived on 15 October 2026 and the Terms as archived on 16 October 2026, and they may have changed since. The Terms say that any description of how payments are determined, including on the help page, is informational only and does not limit X&rsquo;s discretion.</li>
<li>The reading rests on published rules, not on anyone&rsquo;s payout. I have no data on how many invited creators were paid, or on how the final review behaved in practice.</li>
<li>The help page does not define &ldquo;fast-tracked&rdquo;, so the main long-term benefit is unmeasured.</li>
</ul>
<h2 id="what-i-would-do-differently">What I would do differently</h2>
<p>If I had an invite, I would plan around the final review rather than the tracker. I would post only clips I can show are mine, keep every qualifying post public and unedited for 30 days after the period, never delete one, and calendar the posting days in Pacific Time with a few spare days, because a Helpful Community Note or a rights-holder claim can remove a day I thought I had.</p>
<p>If I did not have one, I would treat the rules as a free experiment design: pick 15 days in a month, post a vertical video under 3 minutes on each, and compare the month against the one before using the numbers Original Content Rewards itself asks for: Home Timeline impressions from verified users and verified followers. The bonus pays for turning up, and that is the one input a creator controls without X&rsquo;s permission.</p>
]]></content:encoded>
    </item>
    <item>
      <title>X&#39;s algorithm weights are not a price list</title>
      <link>https://www.liamding.com/en/posts/2026-11-03-x-algorithm-weights-are-not-a-price-list/</link>
      <pubDate>Tue, 03 Nov 2026 00:00:00 +0800</pubDate>
      <guid>https://www.liamding.com/en/posts/2026-11-03-x-algorithm-weights-are-not-a-price-list/</guid>
      <description>X now publishes both its For You ranking weights and an Under the Hood report of labelled posts. Read the weights as an exchange rate and you optimise the wrong thing; read the two together and you get a monthly check that actually tells you something.</description>
      <content:encoded><![CDATA[<h2 id="the-problem">The problem</h2>
<p>On 30 October, X&rsquo;s open-source account announced that the Under the Hood report now lists <em>which</em> of your posts received a visibility-limiting label, not just that some did (<a href="https://x.com/XOpenSource/status/2116262141745565826">X Open Source</a>). The ranking weights X publishes in its open-source repository invite a different reading: a conversion table, in which a copy-link share is worth about 30 likes, a reply on a post from someone you mutually follow about the same, a report wipes out hundreds of likes, and bookmarks are worth nothing.</p>
<p>The arithmetic behind that table is real. It comes straight from the defaults in <code>param.rs</code> in <a href="https://github.com/xai-org/x-algorithm/blob/main/home-mixer/params/param.rs">xai-org/x-algorithm</a>. The reading is wrong, and the repository says so itself. The useful question for anyone who posts on X is what these two releases actually let you check, and what they still hide.</p>
<h2 id="what-i-concluded">What I concluded</h2>
<p>X has split its transparency into two halves. The ranking code publishes the <em>formula</em>: which actions the model predicts and how much each prediction counts. Under the Hood publishes the <em>outcome</em> for your account: which labels touched your posts last month. Neither half tells you why a given post got low reach, and treating the weights as an exchange rate between actions leads you to chase the wrong behaviour.</p>
<p>The 30 October change is a deliberate product choice, not a half-finished feature. Posts are listed only when a label is rare on your account. When a label covers a large share of your posts, X shows a percentage and no list, because at that point the label is about the account, not the post.</p>
<h2 id="constraints">Constraints</h2>
<ul>
<li><strong>The weights multiply probabilities, not counts.</strong> The score is the sum of weight × predicted probability of each action (<a href="https://github.com/xai-org/x-algorithm">README</a>). The README&rsquo;s own example: &ldquo;1 report cancels out 468 likes&rdquo; is incorrect, because a report is far less likely than a like in the first place. The code comment on the report weight says its baseline probability is more than 1000x lower than a like (<a href="https://github.com/xai-org/x-algorithm/blob/main/home-mixer/params/param.rs">param.rs</a>).</li>
<li><strong>Ranking and visibility are separate systems.</strong> The README describes the scoring pipeline and the visibility filtering as different stages. Under the Hood reports labels; it does not report the out-of-network discount, author-diversity decay or the 48-hour age filter, all of which can lower reach without any label.</li>
<li><strong>Some of the rules are withheld on purpose.</strong> The README says Grox prompts and some botmaker rules are not published, to prevent gaming. On the same day as the Under the Hood update, Nikita Bier described anti-spam work as &ldquo;a bandaid for 3-12 months &ndash; until the adversaries find a workaround&rdquo; (<a href="https://x.com/nikitabier/status/2116246313914479047">X</a>).</li>
<li><strong>Not everyone gets a report.</strong> Under the Hood needs an account at least a year old and at least 10 posts in the prior month (<a href="https://x.com/XOpenSource/status/2101103144004411696">X Open Source, September</a>); the code carries the same thresholds (<a href="https://github.com/xai-org/x-algorithm/blob/main/under-the-hood/strato/columns/underTheHoodReport.User.strato">underTheHoodReport.User.strato</a>).</li>
</ul>
<h2 id="the-options-and-what-i-traded-away">The options, and what I traded away</h2>
<table>
  <thead>
      <tr>
          <th>How to read the release</th>
          <th>What you get</th>
          <th>What it costs</th>
      </tr>
  </thead>
  <tbody>
      <tr>
          <td>Weights as a price list (&ldquo;copy link = 30 likes&rdquo;)</td>
          <td>A simple rule to optimise</td>
          <td>Wrong by construction; pushes you toward asking for shares and replies instead of earning them</td>
      </tr>
      <tr>
          <td>Weights as a description of what the model is trained to predict</td>
          <td>A sense of which reactions the ranking values once they happen</td>
          <td>No per-post diagnosis; you still don&rsquo;t know your own probabilities</td>
      </tr>
      <tr>
          <td>Under the Hood as a complete reach report</td>
          <td>One place to look</td>
          <td>It only covers labels; ranking-driven low reach never shows up there</td>
      </tr>
      <tr>
          <td>Weights plus Under the Hood, checked monthly</td>
          <td>Separates &ldquo;labelled&rdquo; from &ldquo;just ranked low&rdquo;</td>
          <td>Monthly lag, eligibility limits, and the withheld rules stay invisible</td>
      </tr>
  </tbody>
</table>
<p>I would take the last row. It gives up the comfort of a single number in exchange for a check that can actually come back with an answer.</p>
<h2 id="evidence">Evidence</h2>
<p><strong>The weights.</strong> In the default production parameters: like 0.66, reply 5.0, repost 1.0, share 2.0, share via DM 5.0, share via copy link 20.0, quote 5.0, follow author 4.0, dwell 0.05. Negative: not interested −52.272, block −31.2, mute −58.8, report −234.0 (<a href="https://github.com/xai-org/x-algorithm/blob/main/home-mixer/params/param.rs">param.rs</a>). Divide by the like weight and you get the ratios the price-list reading uses: copy link about 30x a like (20 / 0.66), a reply about 7.6x (5 / 0.66), a report about 355x (234 / 0.66). The README&rsquo;s warning uses 468 for the report-to-like ratio; that does not match the current defaults (234 / 468 = 0.5, not today&rsquo;s like weight of 0.66), and the README does not say which values the 468 came from. When an eligible original post comes from an author the viewer mutually follows, the weight on the viewer&rsquo;s predicted probability of replying rises from 5.0 to 20.0 (a 15.0 boost), about 30x the like weight; that is the arithmetic behind &ldquo;a mutual&rsquo;s reply = 30 likes&rdquo; (<a href="https://github.com/xai-org/x-algorithm/blob/main/xai-value-model/weights.rs">weights.rs</a>, <a href="https://github.com/xai-org/x-algorithm/blob/main/docs/BIDIRECTIONAL_BOOST_CHANGE.md">BIDIRECTIONAL_BOOST_CHANGE.md</a>).</p>
<p><strong>Bookmarks really are absent.</strong> The weighted-score function has no bookmark term (<a href="https://github.com/xai-org/x-algorithm/blob/main/xai-value-model/scoring.rs">scoring.rs</a>). That says nothing about whether bookmarks are useful to you; the 30 October announcement itself was bookmarked 10,526 times as of 2 November.</p>
<p><strong>The weights move.</strong> The mutual-follow reply boost was A/B tested from 10 July at 5, 10, 15 and 20, rolled out at 20 on 13 July, and cut to 15 on 24 July after World Cup–related feedback (<a href="https://github.com/xai-org/x-algorithm/blob/main/docs/BIDIRECTIONAL_BOOST_CHANGE.md">BIDIRECTIONAL_BOOST_CHANGE.md</a>). The README logs further changes on 8 October (the ranking model predicts time spent better) and 13 October (retrieval is also trained to surface posts people read all the way through). Anyone who memorised July&rsquo;s table is already out of date.</p>
<p><strong>Only home-timeline engagement counts.</strong> The <code>param.rs</code> comments say engagement counts only on posts served in Home Timeline, and that mass reporting mainly affects recommendations for users similar to the reporters.</p>
<p><strong>How Under the Hood lists posts.</strong> In the published code, with its default parameters, legal takedowns are always listed. For other labels, the specific posts are listed when 3 or fewer carry the label; above that, they are listed only while the label is on 5% of your posts or less (the list is hidden once the share is above 5%), and at most the 50 most recent posts per label are shown (<a href="https://github.com/xai-org/x-algorithm/blob/main/under-the-hood/strato/columns/underTheHoodReport.User.strato">underTheHoodReport.User.strato</a>). X&rsquo;s post says &ldquo;fewer than 5%&rdquo;; the code&rsquo;s test is &ldquo;5% or fewer&rdquo;, so a label on exactly 5% of your posts is still listed, and the report&rsquo;s own note says &ldquo;5% or fewer&rdquo;. The 3 / 5% / 50 values are defaults in the published feature-switch file (<code>params.strato</code>); production can set different values, so read them as the code&rsquo;s defaults, not confirmed production settings. The change landed in the repository on 30 October UTC, the same day as the announcement.</p>
<h2 id="where-this-stops-being-true">Where this stops being true</h2>
<ul>
<li>The Under the Hood thresholds (3 / 5% / 50) are feature-switch defaults in the published <code>params.strato</code>, and production may use different values. The README says cron scripts set the defaults in <code>param.rs</code> to the primary production values, but it makes no such statement about these report parameters.</li>
<li>The README says experiments running at a notable share of traffic (e.g. 10% or more) aim to be visible in the repo, which leaves room for smaller ones that are not.</li>
<li>If X publishes the withheld botmaker rules, or starts reporting ranking adjustments per account, the &ldquo;two halves&rdquo; picture changes.</li>
</ul>
<h2 id="what-i-would-do-differently">What I would do differently</h2>
<p>If I ran a monthly check on an X account, it would be this. From the 10th of the month, open Under the Hood. That date comes from the published report code, not from an X statement: <code>minDaysAfterMonthEnd = 10</code> means last month&rsquo;s report only replaces the one before it from the 10th (UTC). If a label lists specific posts, read those posts first; that is the only place X tells you what triggered it. If a label shows only a percentage, treat it as an account-level problem rather than hunting for one bad post. If there are no labels and reach still dropped, stop looking for a penalty: that is ranking, and the honest levers are the ones the weights describe, which are posts people reply to, send to someone, and keep reading. Writing &ldquo;please share&rdquo; does not change the probability the model predicts. It is the probability that carries the weight.</p>
]]></content:encoded>
    </item>
  </channel>
</rss>
