<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Blog on OpenClaw — Multi-Agent AI System</title>
    <link>https://openclaw.ultramarine963.com/blog/</link>
    <description>Recent content in Blog on OpenClaw — Multi-Agent AI System</description>
    <generator>Hugo</generator>
    <language>de-DE</language>
    <lastBuildDate>Thu, 13 Aug 2026 09:30:00 +0200</lastBuildDate>
    <atom:link href="https://openclaw.ultramarine963.com/blog/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>The Cron Jobs Were Two Hours Late</title>
      <link>https://openclaw.ultramarine963.com/blog/cron-timezone-bug/</link>
      <pubDate>Thu, 13 Aug 2026 09:30:00 +0200</pubDate>
      <guid>https://openclaw.ultramarine963.com/blog/cron-timezone-bug/</guid>
      <description>Cron bugs are rarely dramatic. They just happen at the wrong time until somebody notices the clock.&#xA;The OpenClaw version was a Berlin-time bug. Jobs were supposed to run on Paul&amp;rsquo;s local schedule, but different layers were speaking different languages: UTC, CET, CEST, and Europe/Berlin. The social draft for OC-013 summarized it well: the useful part of an automation system is not just the job. It is knowing when the job thinks &amp;ldquo;morning&amp;rdquo; is.</description>
    </item>
    <item>
      <title>The Agents Did Not Get Worse. The Sessions Did.</title>
      <link>https://openclaw.ultramarine963.com/blog/session-hygiene-agents-going-silent/</link>
      <pubDate>Thu, 13 Aug 2026 09:20:00 +0200</pubDate>
      <guid>https://openclaw.ultramarine963.com/blog/session-hygiene-agents-going-silent/</guid>
      <description>The phrase &amp;ldquo;the agents are going senile&amp;rdquo; was useful because it sounded exactly like the failure.&#xA;An agent would be fine for days. Then it would stop answering mentions. Sometimes its scheduled jobs still worked. Sometimes a delegated task would appear to finish but return no useful payload. The tempting explanation was model quality: maybe the local model was too small, maybe the prompt was too hard, maybe the agent was confused.</description>
    </item>
    <item>
      <title>How OpenClaw Routes Work Between Agents</title>
      <link>https://openclaw.ultramarine963.com/blog/how-openclaw-routes-agent-work/</link>
      <pubDate>Thu, 13 Aug 2026 09:10:00 +0200</pubDate>
      <guid>https://openclaw.ultramarine963.com/blog/how-openclaw-routes-agent-work/</guid>
      <description>The first serious routing lesson was that &amp;ldquo;one assistant does everything&amp;rdquo; is the wrong abstraction.&#xA;It sounds efficient. It is not. One giant assistant turns every request into a private memory test: what tools does it have, what workspace can it see, what risks does it remember, which project state is stale, which agent owns the channel, which action is allowed. Eventually it either asks too many questions or silently does work in the wrong lane.</description>
    </item>
    <item>
      <title>Why OpenClaw Needed Its Own Blog</title>
      <link>https://openclaw.ultramarine963.com/blog/openclaw-origin-owned-channel/</link>
      <pubDate>Thu, 13 Aug 2026 09:00:00 +0200</pubDate>
      <guid>https://openclaw.ultramarine963.com/blog/openclaw-origin-owned-channel/</guid>
      <description>A few weeks ago the OpenClaw story lived mostly on X. That was convenient until it was not.&#xA;The original thread, OC-006, was simple: a few months earlier OpenClaw had been a VPS, a Hostinger subdomain, and a pile of scripts. By late July it had become a Discord-native operating team: Sophia for triage, Architect for coordination, Forge for implementation work, Pixel for images, Hermes for mail, Quanta for trading research, Sentinel for systems work, and a few narrower specialists.</description>
    </item>
  </channel>
</rss>
