<?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>Engineering-Practice on OpenAdapt Blog</title><link>https://blog.openadapt.ai/tags/engineering-practice/</link><description>Recent content in Engineering-Practice on OpenAdapt Blog</description><generator>Hugo -- 0.165.0</generator><language>en-us</language><lastBuildDate>Fri, 28 Aug 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://blog.openadapt.ai/tags/engineering-practice/index.xml" rel="self" type="application/rss+xml"/><item><title>The One Number That Drifted Was the One That Changed: Why We Started Checksumming Our Own Benchmark Claims</title><link>https://blog.openadapt.ai/posts/2026-08-28-the-one-number-that-drifted-was-the-one-that-changed-why-we/</link><pubDate>Fri, 28 Aug 2026 00:00:00 +0000</pubDate><guid>https://blog.openadapt.ai/posts/2026-08-28-the-one-number-that-drifted-was-the-one-that-changed-why-we/</guid><description>An audit found that 32 of 33 numeric fields in a published benchmark file matched their upstream source exactly, and the one field that had changed upstream was the one that had drifted, which is why we now checksum every published figure against the artifact it was copied from.</description></item></channel></rss>