<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>AI 2026 on Rajat Sapkota</title><link>https://blog.sapkotarajat.com.np/tags/ai-2026/</link><description>Recent content in AI 2026 on Rajat Sapkota</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 12 Apr 2026 00:50:16 +0000</lastBuildDate><atom:link href="https://blog.sapkotarajat.com.np/tags/ai-2026/index.xml" rel="self" type="application/rss+xml"/><item><title>GPT-5.4 Just Beat Humans. Here's What Comes Next.</title><link>https://blog.sapkotarajat.com.np/posts/gpt-54-just-beat-humans-heres-what-comes-next/</link><pubDate>Sun, 12 Apr 2026 00:50:16 +0000</pubDate><guid>https://blog.sapkotarajat.com.np/posts/gpt-54-just-beat-humans-heres-what-comes-next/</guid><description>OpenAI&amp;#39;s GPT-5.4 &amp;#39;Thinking&amp;#39; model just scored 83% on the GDPVal benchmark — matching human experts on real economic tasks. But the real story isn&amp;#39;t the model. It&amp;#39;s the trillion-dollar compute war happ</description></item></channel></rss>