<?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>Continuous-Learning on ClawSouls Blog</title><link>https://blog.clawsouls.ai/en/tags/continuous-learning/</link><description>Recent content in Continuous-Learning on ClawSouls Blog</description><generator>Hugo -- 0.146.7</generator><language>en</language><lastBuildDate>Fri, 24 Jul 2026 19:10:00 +0900</lastBuildDate><atom:link href="https://blog.clawsouls.ai/en/tags/continuous-learning/index.xml" rel="self" type="application/rss+xml"/><item><title>Continuous Learning Won't Come From the Weights</title><link>https://blog.clawsouls.ai/en/posts/continuous-learning-file-layer/</link><pubDate>Fri, 24 Jul 2026 19:10:00 +0900</pubDate><guid>https://blog.clawsouls.ai/en/posts/continuous-learning-file-layer/</guid><description>DeepSeek&amp;#39;s founder named continuous learning as the single biggest gap on the road to AGI. He&amp;#39;s right — but the answer isn&amp;#39;t retraining the model. It&amp;#39;s a memory layer built in the open, at the file level.</description></item></channel></rss>