<?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>Data-Engineering on Backpressured</title><link>https://backpressured.dev/en/tags/data-engineering/</link><description>Recent content in Data-Engineering on Backpressured</description><generator>Hugo</generator><language>en</language><lastBuildDate>Sat, 12 Sep 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://backpressured.dev/en/tags/data-engineering/index.xml" rel="self" type="application/rss+xml"/><item><title>Do Document Links Help AI Find What It Needs?</title><link>https://backpressured.dev/en/posts/does-linking-knowledge-help-ai/</link><pubDate>Sat, 12 Sep 2026 00:00:00 +0000</pubDate><guid>https://backpressured.dev/en/posts/does-linking-knowledge-help-ai/</guid><description>Can following document links help AI find more relevant documents and use fewer tokens than search? I compared rg search and link navigation using 43 operational documents and 15 questions.</description></item><item><title>Debezium Does Not Replicate the Source As-Is: Three Ways Data Can Drift Quietly</title><link>https://backpressured.dev/en/posts/debezium-cdc-silent-data-errors/</link><pubDate>Sat, 05 Sep 2026 00:00:00 +0000</pubDate><guid>https://backpressured.dev/en/posts/debezium-cdc-silent-data-errors/</guid><description>A CDC pipeline can run without errors while the data it produces no longer has the meaning or shape you expected. These are three quiet data mismatches we ran into while operating Debezium in production.</description></item></channel></rss>