<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>False Systems — Writing</title>
    <link>https://false-systems.com/blog.html</link>
    <atom:link href="https://false-systems.com/feed.xml" rel="self" type="application/rss+xml" />
    <description>Essays on agent authority, evidence, and the infrastructure that has to prove what an autonomous system did.</description>
    <language>en</language>
    <lastBuildDate>Thu, 30 Jul 2026 00:00:00 +0000</lastBuildDate>
    <item>
      <title>A log cannot carry a negative</title>
      <link>https://false-systems.com/a-log-cannot-carry-a-negative.html</link>
      <guid isPermaLink="true">https://false-systems.com/a-log-cannot-carry-a-negative.html</guid>
      <description>Hugging Face published a forensic account of an autonomous agent intrusion. The interesting part is the sentences that cannot be checked.</description>
      <pubDate>Thu, 30 Jul 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
    <item>
      <title>We needed somewhere to put the evidence</title>
      <link>https://false-systems.com/we-needed-somewhere-to-put-the-evidence.html</link>
      <guid isPermaLink="true">https://false-systems.com/we-needed-somewhere-to-put-the-evidence.html</guid>
      <description>Infrastructure doesn't forget — it remembers in fragments, scattered across logs, audit trails, and CI. Why Ahti exists as an append-first evidence database, and why it refuses to interpret what it stores.</description>
      <pubDate>Wed, 01 Jul 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
    <item>
      <title>Non-determinism is not a failure mode</title>
      <link>https://false-systems.com/non-determinism-is-not-a-failure-mode.html</link>
      <guid isPermaLink="true">https://false-systems.com/non-determinism-is-not-a-failure-mode.html</guid>
      <description>The execution layer of the software supply chain was built for same input, same output. Once models enter the loop, non-determinism becomes a property the graph has to model.</description>
      <pubDate>Sun, 03 May 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
    <item>
      <title>Pipelines don’t rot. Their abstractions do.</title>
      <link>https://false-systems.com/pipelines-dont-rot.html</link>
      <guid isPermaLink="true">https://false-systems.com/pipelines-dont-rot.html</guid>
      <description>LangChain pipelines written in 2023 don't run in 2026. The code is fine — the abstractions underneath aren't. What an LLM-ops layer needs, and why.</description>
      <pubDate>Wed, 29 Apr 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
    <item>
      <title>Correlation is a data structure</title>
      <link>https://false-systems.com/correlation-is-a-data-structure.html</link>
      <guid isPermaLink="true">https://false-systems.com/correlation-is-a-data-structure.html</guid>
      <description>Observability was built for humans. When the reader is an AI agent, the whole five-unrelated-streams model falls apart. What replaces it is causality stored as data, not inferred at query time.</description>
      <pubDate>Fri, 17 Apr 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
    <item>
      <title>The same brain wrote both sides</title>
      <link>https://false-systems.com/luotain-blackbox-qa-for-ai-agents.html</link>
      <guid isPermaLink="true">https://false-systems.com/luotain-blackbox-qa-for-ai-agents.html</guid>
      <description>An agent that writes both the code and its test has closed the loop before it starts. What it takes to test something from the outside instead.</description>
      <pubDate>Sat, 28 Mar 2026 00:00:00 +0000</pubDate>
      <author>yair@false-systems.com (Yair Etziony)</author>
    </item>
  </channel>
</rss>
