<?xml version="1.0"?><rss version="2.0"><channel><title>AlgoWatch</title><link>https://algowatchx.pages.dev/</link><description>Plain-English changes to X's algorithm</description><item><title>Ad changes only</title><link>https://algowatchx.pages.dev/changes/2026-10-29.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-29.html</guid><pubDate>Thu, 29 Oct 2026 18:00:00 +0000</pubDate><description>Four new pieces of code decide how ads are spaced between posts. They don&#x27;t change your post&#x27;s score.</description></item><item><title>New-user filter widened to 7 days, and new experimental re-rankers</title><link>https://algowatchx.pages.dev/changes/2026-10-28.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-28.html</guid><pubDate>Wed, 28 Oct 2026 18:00:00 +0000</pubDate><description>This went from 1 day to 7 days.</description></item><item><title>Ranking model starts receiving &#x27;de-bias&#x27; inputs</title><link>https://algowatchx.pages.dev/changes/2026-10-27.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-27.html</guid><pubDate>Tue, 27 Oct 2026 18:00:00 +0000</pubDate><description>The ranking model is now sent extra inputs to correct for bias, likely position bias (posts higher up get more clicks just for being there). The effect on reach is unclear.</description></item><item><title>No ranking changes today</title><link>https://algowatchx.pages.dev/changes/2026-10-24.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-24.html</guid><pubDate>Sat, 24 Oct 2026 18:00:00 +0000</pubDate><description>The only live switch was an ad-tracing tool.</description></item><item><title>Experiment shuffle, nothing visible</title><link>https://algowatchx.pages.dev/changes/2026-10-23.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-23.html</guid><pubDate>Fri, 23 Oct 2026 18:00:00 +0000</pubDate><description>X changed which experimental models get a slice of traffic. We can&#x27;t see results.</description></item><item><title>Fair share ranking switched on, and new users see less racy content</title><link>https://algowatchx.pages.dev/changes/2026-10-22.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-22.html</guid><pubDate>Thu, 22 Oct 2026 18:00:00 +0000</pubDate><description>We think this stops a few authors dominating a feed. That&#x27;s an inference from the name and code, not confirmed.</description></item><item><title>&#x27;Fair share ranking&#x27; is being built</title><link>https://algowatchx.pages.dev/changes/2026-10-21.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-21.html</guid><pubDate>Wed, 21 Oct 2026 18:00:00 +0000</pubDate><description>From the name, it spreads feed space more evenly, probably across authors. It was switched on the next day.</description></item><item><title>Quiet day for creators</title><link>https://algowatchx.pages.dev/changes/2026-10-20.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-20.html</guid><pubDate>Tue, 20 Oct 2026 18:00:00 +0000</pubDate><description>A new setting shows X looks at the last 28 days of your history to pick topics. It&#x27;s new as a setting, not necessarily new behaviour.</description></item><item><title>Video carousel stops showing the same creator over and over</title><link>https://algowatchx.pages.dev/changes/2026-10-17.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-17.html</guid><pubDate>Sat, 17 Oct 2026 18:00:00 +0000</pubDate><description>The carousel now de-duplicates by author and by visual similarity, and both are on by default.</description></item><item><title>More video plumbing</title><link>https://algowatchx.pages.dev/changes/2026-10-16.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-16.html</guid><pubDate>Fri, 16 Oct 2026 18:00:00 +0000</pubDate><description>It added &#x27;video embeddings&#x27;, which are fingerprints of what a video looks like. This lays the groundwork for the next day&#x27;s change.</description></item><item><title>A new &#x27;immersive video&#x27; scorer arrives</title><link>https://algowatchx.pages.dev/changes/2026-10-15.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-15.html</guid><pubDate>Thu, 15 Oct 2026 18:00:00 +0000</pubDate><description>A new scorer for the full-screen video feed was added, weighted on video opens for now. Whether it&#x27;s in heavy use isn&#x27;t clear from the code.</description></item><item><title>Likes up again, and &#x27;not interested&#x27; hurts more</title><link>https://algowatchx.pages.dev/changes/2026-10-14.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-14.html</guid><pubDate>Wed, 14 Oct 2026 18:00:00 +0000</pubDate><description>The like weight went from 0.6 to 0.66, which is +32% across two days.</description></item><item><title>Big one: likes and reading time count for more, video taps count for less</title><link>https://algowatchx.pages.dev/changes/2026-10-13.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-13.html</guid><pubDate>Tue, 13 Oct 2026 18:00:00 +0000</pubDate><description>The weight on likes went from 0.5 to 0.6. Likes push a post further than they did yesterday.</description></item><item><title>No ranking changes today</title><link>https://algowatchx.pages.dev/changes/2026-10-09.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-09.html</guid><pubDate>Fri, 09 Oct 2026 18:00:00 +0000</pubDate><description>Mostly model-training code and data plumbing. Nothing that moves your reach.</description></item><item><title>Reply-spam rules ease off for mid-sized accounts</title><link>https://algowatchx.pages.dev/changes/2026-10-08.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-08.html</guid><pubDate>Thu, 08 Oct 2026 18:00:00 +0000</pubDate><description>That check now only runs on posts from accounts over 180,000 followers (it was 125,000). Big but not huge accounts get looser checks on the replies under their posts.</description></item><item><title>&#x27;Popular posts&#x27; goes live, and fresh posts get a smaller kick</title><link>https://algowatchx.pages.dev/changes/2026-10-07.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-07.html</guid><pubDate>Wed, 07 Oct 2026 18:00:00 +0000</pubDate><description>The popular-posts source from 3 Oct is now switched on by default. Already-popular posts from top authors get an extra route into people&#x27;s feeds.</description></item><item><title>Behind-the-scenes plumbing</title><link>https://algowatchx.pages.dev/changes/2026-10-06.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-06.html</guid><pubDate>Tue, 06 Oct 2026 18:00:00 +0000</pubDate><description>One pipeline now passes up to 100,000 candidates instead of 200. That&#x27;s mostly an engineering change, and what it does to reach is unclear.</description></item><item><title>Groundwork for &#x27;popular posts&#x27; and a small reply-spam tweak</title><link>https://algowatchx.pages.dev/changes/2026-10-03.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-03.html</guid><pubDate>Sat, 03 Oct 2026 18:00:00 +0000</pubDate><description>X built a source that pulls in popular posts from the top 1,000 authors, max 3 each. It was switched on four days later (see 7 Oct).</description></item><item><title>X is building a new way to find posts for you</title><link>https://algowatchx.pages.dev/changes/2026-10-02.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-02.html</guid><pubDate>Fri, 02 Oct 2026 18:00:00 +0000</pubDate><description>X added a new place to pull recommended posts from (called &#x27;sid source&#x27;). From the settings it looks like it finds posts similar to ones you&#x27;ve already engaged with, capped at 800 per feed. We can&#x27;t see whether it&#x27;s switched on for everyone yet.</description></item><item><title>Quiet day: a tidy-up to feed surveys</title><link>https://algowatchx.pages.dev/changes/2026-10-01.html</link><guid>https://algowatchx.pages.dev/changes/2026-10-01.html</guid><pubDate>Thu, 01 Oct 2026 18:00:00 +0000</pubDate><description>X swapped how it times the little &#x27;how&#x27;s your feed?&#x27; surveys, from hours to minutes. Same 24-hour gap as before, just counted differently.</description></item></channel></rss>
