<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"
     xmlns:atom="http://www.w3.org/2005/Atom"
     xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>AI 2027 vs Reality — Daily Signals</title>
    <link>https://ai-2027-vs-reality.com/</link>
    <atom:link href="https://ai-2027-vs-reality.com/feed.xml" rel="self" type="application/rss+xml"/>
    <description>A continuously updated journalistic track record of the AI 2027 scenario vs. reality. Daily signals from AI labs, coding-agent capabilities, China compute, alignment research.</description>
    <language>en</language>
    <pubDate>Sat, 18 Apr 2026 09:16:36 GMT</pubDate>
    <lastBuildDate>Sat, 18 Apr 2026 09:16:36 GMT</lastBuildDate>
    <ttl>60</ttl>
    <item>
      <title>DeepSeek V4 to train and serve exclusively on Huawei Ascend 950PR</title>
      <link>https://ai-2027-vs-reality.com/#signal-2026-04-18-03-deepseek-v4-ascend</link>
      <guid isPermaLink="false">https://ai-2027-vs-reality.com/signal/2026-04-18-03-deepseek-v4-ascend</guid>
      <pubDate>Sat, 18 Apr 2026 00:00:00 GMT</pubDate>
      <category>china-compute</category>
      <description><![CDATA[<p>According to reporting aggregated by <a href="https://techwireasia.com/2026/04/deepseek-v4-points-to-growing-use-of-huawei-chips-in-ai-models/">Tech Wire Asia</a> (citing Reuters and The Information), DeepSeek's forthcoming V4 model — a roughly 1T-parameter MoE slated for late-April release — will be trained and served entirely on Huawei Ascend 950PR silicon. DeepSeek is said to have declined early-access offers from NVIDIA and AMD.</p>
<p>Alibaba, ByteDance, and Tencent have reportedly placed bulk Ascend orders, pushing prices up around 20%. The move is the clearest datapoint yet that a Chinese frontier lab is willing to commit a full training and inference stack off CUDA, though the shift appears ecosystem-driven rather than state-directed.</p>
]]></description>
    </item>
    <item>
      <title>Anthropic completes Claude Mythos but withholds public release under &#39;Project Glasswing&#39;</title>
      <link>https://ai-2027-vs-reality.com/#signal-2026-04-18-02-claude-mythos-withheld</link>
      <guid isPermaLink="false">https://ai-2027-vs-reality.com/signal/2026-04-18-02-claude-mythos-withheld</guid>
      <pubDate>Sat, 18 Apr 2026 00:00:00 GMT</pubDate>
      <category>model-release</category>
      <description><![CDATA[<p>Anthropic confirmed Claude Mythos on April 7 as its most capable model to date, reporting 93.9% on SWE-bench Verified and 94.6% on GPQA Diamond, and describing autonomous discovery of thousands of zero-day vulnerabilities during evaluations. The company stated it will not release the model publicly, restricting access to roughly 50 partner organizations under a program it calls <em>Project Glasswing</em>.</p>
<p>This appears to be the first instance of a frontier lab finishing a flagship model and declining to ship it on safety grounds. The AI 2027 scenario places an analogous decision — Agent-2 judged too capable for general release — in January 2027; the observed event precedes that beat by roughly nine months.</p>
]]></description>
    </item>
    <item>
      <title>DeepSeek V4 to ship on Huawei Ascend 950PR; Nvidia CEO calls it a threat to US dominance</title>
      <link>https://ai-2027-vs-reality.com/#signal-2026-04-18-01-deepseek-v4-ascend</link>
      <guid isPermaLink="false">https://ai-2027-vs-reality.com/signal/2026-04-18-01-deepseek-v4-ascend</guid>
      <pubDate>Sat, 18 Apr 2026 00:00:00 GMT</pubDate>
      <category>china-compute</category>
      <description><![CDATA[<p>On April 16, Nvidia CEO Jensen Huang <a href="https://dataconomy.com/2026/04/16/deepseek-v4-ai-model-to-debut-using-only-huawei-chips/">publicly characterized</a> DeepSeek's forthcoming V4 — reportedly the first frontier-scale model trained end-to-end on Huawei Ascend 950PR silicon — as "a big threat to US dominance," arguing it could establish a non-CUDA reference stack outside the reach of US export controls.</p>
<p>The remark follows <a href="https://www.trendforce.com/news/2026/04/07/news-decoding-deepseek-v4-how-huaweis-ascend-950-pr-is-powering-chinas-push-to-break-cuda-dependence/">reports</a> that Alibaba, ByteDance, and Tencent have placed hundreds-of-thousands-unit orders for Ascend 950PR, with Huawei targeting 1.6M total Ascend units in 2026. If accurate, Chinese frontier compute is consolidating through a domestic supply chain and open-weights distribution rather than the centralized state cluster (e.g., "CDZ at Tianwan") posited in AI-2027.</p>
]]></description>
    </item>
  </channel>
</rss>
