<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>China AI Desk</title>
    <link>https://chinaaidesk.com/</link>
    <description>An independent journal of China AI, grounded in primary sources.</description>
    <language>en</language>
    <lastBuildDate>Thu, 30 Jul 2026 00:00:00 GMT</lastBuildDate>
    <atom:link href="https://chinaaidesk.com/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Issue 08: Qwen-Audio 3.0 TTS — a cloned voice is not consent</title>
      <link>https://chinaaidesk.com/issue-08.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-08.html</guid>
      <pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate>
      <description>What Qwen-Audio 3.0 TTS establishes about real-time synthesis and voice cloning, and why a convincing voice still needs consent, provenance, and revocation controls.</description>
    </item>
    <item>
      <title>Issue 07: Qwen Code worktrees — isolation is not authority</title>
      <link>https://chinaaidesk.com/issue-07.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-07.html</guid>
      <pubDate>Mon, 27 Jul 2026 00:00:00 GMT</pubDate>
      <description>What Qwen Code's worktree sessions isolate, what they still share, and why a separate checkout does not authorize a separate agent.</description>
    </item>
    <item>
      <title>Issue 06: RxBrain — an imagined next state is not an executed action</title>
      <link>https://chinaaidesk.com/issue-06.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-06.html</guid>
      <pubDate>Sat, 25 Jul 2026 00:00:00 GMT</pubDate>
      <description>What Tencent's RxBrain release establishes about joint language-visual planning, and what remains unproven between an imagined goal state and robot action.</description>
    </item>
    <item>
      <title>Issue 05: GLM-5.2 — open weights do not make a 744B system local</title>
      <link>https://chinaaidesk.com/issue-05.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-05.html</guid>
      <pubDate>Fri, 24 Jul 2026 00:00:00 GMT</pubDate>
      <description>What GLM-5.2's open artifacts establish, and what a team still has to prove before calling a 744B-class model local.</description>
    </item>
    <item>
      <title>Issue 04: Qwen Code — learning a workflow is not verifying it</title>
      <link>https://chinaaidesk.com/issue-04.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-04.html</guid>
      <pubDate>Thu, 23 Jul 2026 00:00:00 GMT</pubDate>
      <description>What Qwen Code's learned Skills and review-loop update establishes, and the evidence a team still needs before relying on an automated workflow.</description>
    </item>
    <item>
      <title>Issue 03: DeepSeek V4 — the alias is not the migration</title>
      <link>https://chinaaidesk.com/issue-03.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/issue-03.html</guid>
      <pubDate>Thu, 23 Jul 2026 00:00:00 GMT</pubDate>
      <description>What DeepSeek's legacy model-ID cutoff establishes, what it leaves open, and what an API cutover must verify.</description>
    </item>
    <item>
      <title>Issue 02: Kimi K3 — product access first, weights later</title>
      <link>https://chinaaidesk.com/model.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/model.html</guid>
      <pubDate>Wed, 22 Jul 2026 00:00:00 GMT</pubDate>
      <description>A time-bounded reading of what the Kimi K3 source record supported and what remained for later.</description>
    </item>
    <item>
      <title>Issue 01: The ledger before the voucher</title>
      <link>https://chinaaidesk.com/policy.html</link>
      <guid isPermaLink="true">https://chinaaidesk.com/policy.html</guid>
      <pubDate>Wed, 22 Jul 2026 00:00:00 GMT</pubDate>
      <description>A source-bounded reading of Shanghai's 2026 computing, model, and corpus voucher application notice.</description>
    </item>
  </channel>
</rss>
