<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0">
    <channel>
      <title>QinCloud · Learnings</title>
      <link>https://notes.sparboard.com</link>
      <description>Last 10 notes on QinCloud · Learnings</description>
      <generator>Quartz -- quartz.jzhao.xyz</generator>
      <item>
    <title>Agent Guide</title>
    <link>https://notes.sparboard.com/_meta/agents</link>
    <guid>https://notes.sparboard.com/_meta/agents</guid>
    <description><![CDATA[ Agent Guide How an AI agent should read and extend this learnings/ vault. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Vault Conventions</title>
    <link>https://notes.sparboard.com/_meta/conventions</link>
    <guid>https://notes.sparboard.com/_meta/conventions</guid>
    <description><![CDATA[ Vault Conventions How this learnings/ vault is structured, so it stays consistent as it grows and so both humans and agents can navigate and extend it. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Adversarial Review (Prove It Wrong)</title>
    <link>https://notes.sparboard.com/concepts/adversarial-review</link>
    <guid>https://notes.sparboard.com/concepts/adversarial-review</guid>
    <description><![CDATA[ Adversarial Review (Prove It Wrong) The principle in one line: point a fresh reviewer at your work whose only job is to make it lie or fall over — then check each finding against the running system, because a claim you can test empirically must not be argued. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Fail Loud at the Boundaries</title>
    <link>https://notes.sparboard.com/concepts/fail-loud-at-boundaries</link>
    <guid>https://notes.sparboard.com/concepts/fail-loud-at-boundaries</guid>
    <description><![CDATA[ Fail Loud at the Boundaries The principle in one line: validate at every trust boundary and stop at the cause — a component that silently substitutes a default for bad input isn’t degraded, it’s broken and lying about it. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Idempotent, Self-Verifying Operations</title>
    <link>https://notes.sparboard.com/concepts/idempotent-self-verifying-operations</link>
    <guid>https://notes.sparboard.com/concepts/idempotent-self-verifying-operations</guid>
    <description><![CDATA[ Idempotent, Self-Verifying Operations The principle in one line: an operation should be safe to run again and should prove it worked — never “ran without error, so it must be fine.” What it means (plain English) Two different promises, both essential. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Layered Trust &amp; Defense in Depth</title>
    <link>https://notes.sparboard.com/concepts/layered-trust-defense-in-depth</link>
    <guid>https://notes.sparboard.com/concepts/layered-trust-defense-in-depth</guid>
    <description><![CDATA[ Layered Trust &amp; Defense in Depth The principle in one line: reachability is not authorization — stack several independent controls so that any one of them failing does not open the door. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Make Invalid States Unrepresentable</title>
    <link>https://notes.sparboard.com/concepts/make-invalid-states-unrepresentable</link>
    <guid>https://notes.sparboard.com/concepts/make-invalid-states-unrepresentable</guid>
    <description><![CDATA[ Make Invalid States Unrepresentable The principle in one line: don’t detect-and-repair a bad state after it happens — shape your types and schema so the bad state cannot be written down in the first place. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Observe What Matters (and Prove the Signal)</title>
    <link>https://notes.sparboard.com/concepts/observe-what-matters</link>
    <guid>https://notes.sparboard.com/concepts/observe-what-matters</guid>
    <description><![CDATA[ Observe What Matters (and Prove the Signal) The principle in one line: a metric or alert isn’t done until you’ve seen it fire correctly, and it must be labelled the way a human asks the question. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>Root Cause Over Patch</title>
    <link>https://notes.sparboard.com/concepts/root-cause-over-patch</link>
    <guid>https://notes.sparboard.com/concepts/root-cause-over-patch</guid>
    <description><![CDATA[ Root Cause Over Patch The principle in one line: a visible failure is a pointer to a real defect — keep asking “why was this even possible?” until you reach the origin, then fix that so the whole class of bug cannot recur. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item><item>
    <title>One Source of Truth per Fact</title>
    <link>https://notes.sparboard.com/concepts/single-source-of-truth</link>
    <guid>https://notes.sparboard.com/concepts/single-source-of-truth</guid>
    <description><![CDATA[ One Source of Truth per Fact The principle in one line: every fact has exactly one authority that owns it — everything else derives from that authority and never keeps its own copy. ]]></description>
    <pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate>
  </item>
    </channel>
  </rss>