<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Semgram blog</title><description>What we learn building an AI detector, and what it means for your writing.</description><link>https://semgram.com/</link><item><title>Hidden characters in copied text</title><link>https://semgram.com/blog/the-marks-hidden-in-a-text/</link><guid isPermaLink="true">https://semgram.com/blog/the-marks-hidden-in-a-text/</guid><description>Text you copy out of a chatbot, a PDF or a web page often carries characters you can&apos;t see. What they are, and how to find them.</description><pubDate>Tue, 27 Oct 2026 00:00:00 GMT</pubDate></item><item><title>What a false-positive rate means for your writing</title><link>https://semgram.com/blog/what-a-false-positive-rate-means/</link><guid isPermaLink="true">https://semgram.com/blog/what-a-false-positive-rate-means/</guid><description>Every AI detector flags some human writing by mistake, and how often it does is the number worth asking for.</description><pubDate>Tue, 27 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Why we never tune Detect against Humanize</title><link>https://semgram.com/blog/why-we-never-tune-detect-against-humanize/</link><guid isPermaLink="true">https://semgram.com/blog/why-we-never-tune-detect-against-humanize/</guid><description>We&apos;re building a detector and a humanizer, so we test the humanizer the way we&apos;d test any other writer.</description><pubDate>Tue, 27 Oct 2026 00:00:00 GMT</pubDate></item></channel></rss>