<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
  <title>Cairn — Blog</title>
  <subtitle>Local-first architecture, browser-extension security, AI data containment.</subtitle>
  <link rel="self" type="application/atom+xml" href="https://cairn.blacktrace.co/blog/feed.xml"/>
  <link rel="alternate" type="text/html" href="https://cairn.blacktrace.co/blog"/>
  <id>https://cairn.blacktrace.co/blog/</id>
  <updated>2026-05-29T18:00:00Z</updated>
  <author>
    <name>Blacktrace</name>
    <uri>https://blacktrace.co</uri>
    <email>trace@blacktrace.co</email>
  </author>
  <icon>https://cairn.blacktrace.co/static/favicon.ico</icon>
  <logo>https://blacktrace.co/static/blacktrace-og.jpg</logo>
  <rights>© Blacktrace</rights>

  <entry>
    <title>The folder of zips and grep</title>
    <link rel="alternate" type="text/html" href="https://cairn.blacktrace.co/blog/002-the-folder-of-zips-and-grep"/>
    <id>https://cairn.blacktrace.co/blog/002-the-folder-of-zips-and-grep</id>
    <published>2026-05-29T18:00:00Z</published>
    <updated>2026-05-29T18:00:00Z</updated>
    <author><name>Patrick @ Blacktrace</name></author>
    <summary type="text">Every solo high-volume Claude user already has a folder of exported conversations and a grep command. Cairn is the version of that folder that knows what is in it.</summary>
    <content type="html"><![CDATA[<p>Open the downloads folder of any solo operator who has used Claude seriously for six months. There is a directory. It is called <code>claude-exports</code>, or <code>chats</code>, or it is not called anything and the files live next to a year of PDF receipts. Inside it: zip files. Each one a single conversation, exported by hand because the operator wanted to keep the part where the model finally got it right.</p><p>The folder grows. Nobody curates it. The names are timestamps and uuids. The operator knows three of them by heart and the rest are dark matter. When they need something, they remember a phrase, open a terminal, and run <code>grep -r "the phrase" ~/claude-exports</code>. Sometimes it finds it. Sometimes the phrase was paraphrased and the grep returns nothing and the operator gives up and re-asks the model.</p><p>This is the competitive landscape for Cairn. Not a SaaS dashboard. Not an enterprise platform. A folder of zips and grep.</p><p>Read the full post at <a href="https://cairn.blacktrace.co/blog/002-the-folder-of-zips-and-grep">cairn.blacktrace.co/blog/002-the-folder-of-zips-and-grep</a>.</p>]]></content>
    <category term="cairn"/>
    <category term="positioning"/>
    <category term="solo-operator"/>
  </entry>

  <entry>
    <title>The Shadow-IT Reality of Generative AI in Engineering Teams</title>
    <link rel="alternate" type="text/html" href="https://cairn.blacktrace.co/blog/001-shadow-it-llm-exfiltration"/>
    <id>https://cairn.blacktrace.co/blog/001-shadow-it-llm-exfiltration</id>
    <published>2026-05-29T00:00:00Z</published>
    <updated>2026-05-29T00:00:00Z</updated>
    <author><name>Blacktrace</name></author>
    <summary type="text">Network-level blocking fails because the engineer's incentive to ship outweighs the security team's incentive to gate. This piece examines the actual data path of an LLM paste, the categories of leak it creates, and the architectural property a containment tool needs to have to make the trade survivable.</summary>
    <content type="html">&lt;p&gt;Every engineering organisation of any size has, in 2026, three categories of LLM user: the ones who use a sanctioned account, the ones who use a personal account, and the ones who do both depending on what they are pasting. The third group is the largest.&lt;/p&gt;&lt;p&gt;Read the full post at &lt;a href="https://cairn.blacktrace.co/blog/001-shadow-it-llm-exfiltration"&gt;cairn.blacktrace.co/blog/001-shadow-it-llm-exfiltration&lt;/a&gt;.&lt;/p&gt;</content>
    <category term="security"/>
    <category term="browser-extensions"/>
    <category term="ai"/>
  </entry>
</feed>
