<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Paul Rukwaro — Engineering notes</title><description>Engineering notes on distributed systems, retrieval evaluation and reliability, written from self-directed technical study alongside the project library.</description><link>https://paulrukwaro.com/</link><language>en-us</language><item><title>The boundary is where modernization risk lives</title><link>https://paulrukwaro.com/notes/boundary-is-where-modernization-risk-lives/</link><guid isPermaLink="true">https://paulrukwaro.com/notes/boundary-is-where-modernization-risk-lives/</guid><description>A controlled C++/.NET interop study that separates boundary hardening from compiler settings, so a hostile-input comparison and a compiler comparison each isolate exactly one variable.</description><pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate><category>Modernization interop</category></item><item><title>Diagnosing where a retrieval pipeline loses evidence</title><link>https://paulrukwaro.com/notes/diagnosing-retrieval-evidence-loss/</link><guid isPermaLink="true">https://paulrukwaro.com/notes/diagnosing-retrieval-evidence-loss/</guid><description>An offline evaluation lab that attributes evidence loss to a specific pipeline stage — chunking, retrieval or ranking — using a synthetic corpus with known evidence spans and a predict-then-measure discipline.</description><pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate><category>Retrieval evaluation</category></item><item><title>Detecting AI answer-quality regressions that don&apos;t throw errors</title><link>https://paulrukwaro.com/notes/detecting-ai-quality-regressions-that-dont-throw-errors/</link><guid isPermaLink="true">https://paulrukwaro.com/notes/detecting-ai-quality-regressions-that-dont-throw-errors/</guid><description>A simulation of the gap between request-level monitoring and answer-quality monitoring, comparing aggregate and topic-sliced drift detectors on deterministic, non-semantic features under a modeled operating budget.</description><pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate><category>AI observability</category></item></channel></rss>