<?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>Agent on AI Pulse</title>
    <link>https://cuigh.com/tags/agent/</link>
    <description>Recent content in Agent on AI Pulse</description>
    <generator>Hugo</generator>
    <language>en-US</language>
    <lastBuildDate>Sun, 17 May 2026 10:30:56 +0800</lastBuildDate>
    <atom:link href="https://cuigh.com/tags/agent/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>What If Your Company Is Getting High on AI?</title>
      <link>https://cuigh.com/posts/companies-ai-psychosis/</link>
      <pubDate>Sun, 17 May 2026 10:30:56 +0800</pubDate>
      <guid>https://cuigh.com/posts/companies-ai-psychosis/</guid>
      <description>A Hacker News thread about AI psychosis became a useful warning: companies are overestimating AI agents, underestimating verification costs, and confusing demos with reliable workflows.</description>
    </item>
    <item>
      <title>MCP, Memory, and Skills: Personal Agents Are Finally Growing a Skeleton</title>
      <link>https://cuigh.com/posts/personal-agent-infrastructure-mcp-memory-skills/</link>
      <pubDate>Mon, 11 May 2026 15:57:18 +0800</pubDate>
      <guid>https://cuigh.com/posts/personal-agent-infrastructure-mcp-memory-skills/</guid>
      <description>From the rise of Skills and Memory projects on GitHub to GBrain&amp;#39;s MCP thin client, Claude Managed Agents memory, and tool-routing products on Product Hunt, this essay argues that the next generation of personal AI agents needs three layers of infrastructure: rules, memory, and tool access.</description>
    </item>
    <item>
      <title>Cheap Local AI Tools Are Starting to Replace Closed-Source Options</title>
      <link>https://cuigh.com/posts/cheap-local-ai-tools-are-starting-to-replace-closed-source/</link>
      <pubDate>Sat, 09 May 2026 14:17:35 +0800</pubDate>
      <guid>https://cuigh.com/posts/cheap-local-ai-tools-are-starting-to-replace-closed-source/</guid>
      <description>From DeepSeek-TUI and local-deep-research to 9router, this essay looks at why local AI tools are moving from hacker toys to serious alternatives for individuals and small teams.</description>
    </item>
    <item>
      <title>The AI Agents Most Likely to Make Money Stay Close to Sales and Launch</title>
      <link>https://cuigh.com/posts/business-agents-that-make-money-stay-close-to-sales-and-launch/</link>
      <pubDate>Tue, 28 Apr 2026 09:09:52 +0800</pubDate>
      <guid>https://cuigh.com/posts/business-agents-that-make-money-stay-close-to-sales-and-launch/</guid>
      <description>Using Orange Slice, Jet AI Agents, and Waitlister as examples, this piece argues that the business agents most likely to survive are the ones closest to sales, launches, acquisition, and conversion.</description>
    </item>
    <item>
      <title>The Real Bottleneck Is Not Agents. It Is Work Worth Giving to Them</title>
      <link>https://cuigh.com/posts/agents-are-not-the-bottleneck-work-is/</link>
      <pubDate>Fri, 24 Apr 2026 12:22:22 +0800</pubDate>
      <guid>https://cuigh.com/posts/agents-are-not-the-bottleneck-work-is/</guid>
      <description>Starting from a sharp line by Zara Zhang, this piece connects Workspace Agents, Teams Agent, Zed Parallel Agents, and Every&amp;#39;s employee-agent experiment to argue that what is scarce today is not agents. It is work that is structured, permissioned, and reviewable enough to hand to them.</description>
    </item>
  </channel>
</rss>
