<?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>Opensource on Onur Baydaş</title><link>https://onurbaydas.com/en/tags/opensource/</link><description>Recent content in Opensource on Onur Baydaş</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Mon, 19 Oct 2026 07:41:16 +0300</lastBuildDate><atom:link href="https://onurbaydas.com/en/tags/opensource/index.xml" rel="self" type="application/rss+xml"/><item><title>Google Open Sources Harness2 for Self-Improving AI Agents</title><link>https://onurbaydas.com/en/news/2026-10-19-google-harness2-ai-agents/</link><pubDate>Mon, 19 Oct 2026 07:41:16 +0300</pubDate><guid>https://onurbaydas.com/en/news/2026-10-19-google-harness2-ai-agents/</guid><description>Google has open-sourced Harness2, which transforms any AI &amp;#34;harness&amp;#34;—the components surrounding an AI model—into a self-improving system. This tool can optimize elements like system prompts, tools, memory, and subagents over time.</description></item><item><title>SonicBoom v0.0.3: Compiler-First AI Runtime Outperforming llama.cpp</title><link>https://onurbaydas.com/en/news/2026-10-19-sonicboom-ai-runtime-benchmark/</link><pubDate>Mon, 19 Oct 2026 07:41:16 +0300</pubDate><guid>https://onurbaydas.com/en/news/2026-10-19-sonicboom-ai-runtime-benchmark/</guid><description>The HardenedLinux community has announced SonicBoom v0.0.3, an open-source (GPLv3) inference runtime developed as an alternative to Python-based solutions. This new runtime, which outperforms llama.cpp in decode performance, is seeking GPU benchmarks from the community.</description></item></channel></rss>