<?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>Architecture on mu0</title><link>https://mu0.ai/tags/architecture/</link><description>Recent content in Architecture on mu0</description><generator>Hugo -- 0.152.2</generator><language>en-us</language><lastBuildDate>Tue, 22 Sep 2026 09:00:00 +1000</lastBuildDate><atom:link href="https://mu0.ai/tags/architecture/index.xml" rel="self" type="application/rss+xml"/><item><title>Guessing the Jev Architecture</title><link>https://mu0.ai/posts/guessing-the-jev-architecture/</link><pubDate>Tue, 22 Sep 2026 09:00:00 +1000</pubDate><guid>https://mu0.ai/posts/guessing-the-jev-architecture/</guid><description>TypeSafe has released Jev, a &amp;lsquo;System One&amp;rsquo; model that returns typed, calibrated decisions instead of text. The architecture is not public. Here is the one I would bet on — a bidirectional encoder over the state, plus a language-conditioned cross-attention head that scores every candidate answer in parallel.</description></item></channel></rss>