<?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>Model Routing on DataThrillz</title><link>https://datathrillz.com/tags/model-routing/</link><description>Recent content in Model Routing on DataThrillz</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Thu, 24 Sep 2026 12:00:00 +0000</lastBuildDate><atom:link href="https://datathrillz.com/tags/model-routing/index.xml" rel="self" type="application/rss+xml"/><item><title>TIL: Jev and the Rise of Decision Models</title><link>https://datathrillz.com/posts/2026-09-24-jev-decision-models-til/</link><pubDate>Thu, 24 Sep 2026 12:00:00 +0000</pubDate><guid>https://datathrillz.com/posts/2026-09-24-jev-decision-models-til/</guid><description>&lt;p&gt;Last week, TypeSafe AI introduced &lt;strong&gt;&lt;a href="https://docs.typesafe.ai/introduction.md"&gt;Jev&lt;/a&gt;&lt;/strong&gt;, a new category of model it calls a &lt;em&gt;decision model&lt;/em&gt;. Unlike an LLM, Jev is not designed to generate prose. It takes unstructured text or JSON and returns a typed decision: a category, score, yes/no answer, extracted value, and an associated confidence score.&lt;/p&gt;
&lt;p&gt;TypeSafe describes it as &amp;ldquo;a frontier-intelligence function call: unstructured state in, typed probabilistic decisions out.&amp;rdquo; The important distinction is that a program can branch directly on the result instead of interpreting generated text.&lt;/p&gt;</description></item></channel></rss>