<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Manoj Bajaj</title>
    <link>https://mbajaj.me</link>
    <description>Essays on AI agent infrastructure — identity, credential management, and the missing primitives for safe agent deployment.</description>
    <language>en-us</language>
    <atom:link href="https://mbajaj.me/rss.xml" rel="self" type="application/rss+xml"/>
    
    <item>
      <title><![CDATA[You’re Still Building for the Wrong User]]></title>
      <link>https://mbajaj.me/blog/you-re-still-building-for-the-wrong-user</link>
      <guid>https://mbajaj.me/blog/you-re-still-building-for-the-wrong-user</guid>
      <pubDate>Thu, 28 May 2026 17:05:00 GMT</pubDate>
      <description><![CDATA[]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[The Agent Economy Needs a New Stack]]></title>
      <link>https://mbajaj.me/blog/the-agent-economy-needs-a-new-stack</link>
      <guid>https://mbajaj.me/blog/the-agent-economy-needs-a-new-stack</guid>
      <pubDate>Thu, 28 May 2026 17:04:00 GMT</pubDate>
      <description><![CDATA[]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[Agent Tooling Should Be Composable, Not Convenient]]></title>
      <link>https://mbajaj.me/blog/agent-tooling-should-be-composable-not-convenient</link>
      <guid>https://mbajaj.me/blog/agent-tooling-should-be-composable-not-convenient</guid>
      <pubDate>Thu, 14 May 2026 07:46:00 GMT</pubDate>
      <description><![CDATA[Most software is designed around human fatigue. We reduce clicks, hide options, compress flows, and celebrate when a task goes from ten steps to one. That instinct makes sense for humans. It is not obviously right for agents. Agents can call tools sequentially, write glue code, retry failed steps, inspect intermediate outputs, and compose workflows without getting bored. For agents, the goal is not fewer steps. The goal is better primitives.]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[Agent Identity: Core agent primitive]]></title>
      <link>https://mbajaj.me/blog/agent-identity-core-agent-primitive</link>
      <guid>https://mbajaj.me/blog/agent-identity-core-agent-primitive</guid>
      <pubDate>Fri, 17 Apr 2026 06:04:00 GMT</pubDate>
      <description><![CDATA[Agents are becoming real software actors, not just chat.
Once they act on systems, the core problem shifts from capability to authority: who they are, what they can do, and on whose behalf.
Most tools today solve slices like auth, secrets, or gateways. What’s missing is a clean identity layer that ties it all together:
user → agent → runtime → tool → action]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[Minimal Agent Primitives]]></title>
      <link>https://mbajaj.me/blog/minimal-agent-primitives</link>
      <guid>https://mbajaj.me/blog/minimal-agent-primitives</guid>
      <pubDate>Tue, 24 Mar 2026 16:41:00 GMT</pubDate>
      <description><![CDATA[]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[MCP vs CLI is wrong question]]></title>
      <link>https://mbajaj.me/blog/mcp-vs-cli-is-wrong-question</link>
      <guid>https://mbajaj.me/blog/mcp-vs-cli-is-wrong-question</guid>
      <pubDate>Thu, 12 Mar 2026 09:09:00 GMT</pubDate>
      <description><![CDATA[]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
    <item>
      <title><![CDATA[Don't Build Agents, Build Primitives.]]></title>
      <link>https://mbajaj.me/blog/don-t-build-agents-build-primitives</link>
      <guid>https://mbajaj.me/blog/don-t-build-agents-build-primitives</guid>
      <pubDate>Fri, 06 Mar 2026 13:35:00 GMT</pubDate>
      <description><![CDATA[]]></description>
      <author>manoj@ruzo.ai (Manoj Bajaj)</author>
    </item>
  </channel>
</rss>