<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Harness on Radar · Local AI &amp; Agent Security</title><link>https://radar.clouddomus.com/categories/harness/</link><description>Recent content in Harness on Radar · Local AI &amp; Agent Security</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Tue, 17 Nov 2026 23:40:00 -0300</lastBuildDate><atom:link href="https://radar.clouddomus.com/categories/harness/index.xml" rel="self" type="application/rss+xml"/><item><title>DeepSeek Harness 0.2.2-rc.4 fixes LiveShare and handles unexpected model responses better</title><link>https://radar.clouddomus.com/news/2026-11/dsh-0-2-2-rc4/</link><pubDate>Tue, 17 Nov 2026 23:40:00 -0300</pubDate><guid>https://radar.clouddomus.com/news/2026-11/dsh-0-2-2-rc4/</guid><description>A release candidate that fixes LiveShare session creation in the desktop app and handles out-of-spec responses from the model more gracefully.</description></item><item><title>A LoRA trained with RL inside the Pi harness doubles Qwen3.8-27B on engineering tasks</title><link>https://radar.clouddomus.com/news/2026-11/qwen38-27b-pi-adapted/</link><pubDate>Tue, 17 Nov 2026 17:30:00 -0300</pubDate><guid>https://radar.clouddomus.com/news/2026-11/qwen38-27b-pi-adapted/</guid><description>A rank-32 LoRA, trained by reinforcement learning inside the Pi harness itself, takes the 27B from 29.4% to 62.7% on SWE-bench Pro Hard, according to its authors.</description></item></channel></rss>