<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Konrad Madej</title><description>Konrad Madej writes about enterprise AI, platform engineering, and DevOps — lessons from real-world projects.</description><link>https://konradmadej.com/</link><language>en</language><item><title>Your AI Pair Programmer Has Terrible Taste in Dependencies</title><link>https://konradmadej.com/blog/2026-07-18-your-ai-pair-programmer-has-terrible-taste-in-dependencies/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-07-18-your-ai-pair-programmer-has-terrible-taste-in-dependencies/</guid><description>LLMs pick libraries by remembering, not evaluating: frozen favourites, hallucinated packages, vulnerable version pins. Where the risk comes from and the guardrails that work.</description><pubDate>Sat, 18 Jul 2026 12:00:00 GMT</pubDate><category>AI</category><category>ai</category><category>supply-chain</category><category>dependencies</category><category>security</category><author>Konrad Madej</author></item><item><title>Dependabot Says Update Now. Supply-Chain Attacks Say Slow Down.</title><link>https://konradmadej.com/blog/2026-07-18-dependabot-says-update-now-supply-chain-attacks-say-slow-down/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-07-18-dependabot-says-update-now-supply-chain-attacks-say-slow-down/</guid><description>Updating dependencies promptly is a security practice for known vulnerabilities and an attack surface for unknown ones. The difference between the two is about seven days of patience.</description><pubDate>Sat, 18 Jul 2026 11:00:00 GMT</pubDate><category>DevOps</category><category>supply-chain</category><category>dependencies</category><category>security</category><category>devops</category><author>Konrad Madej</author></item><item><title>Sometimes the Missing Layer Should Stay Missing</title><link>https://konradmadej.com/blog/2026-07-18-sometimes-the-missing-layer-should-stay-missing/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-07-18-sometimes-the-missing-layer-should-stay-missing/</guid><description>The most expensive platform failures aren&apos;t badly built platforms — they&apos;re platforms that should never have been started. How to tell, and the ladder of cheaper alternatives.</description><pubDate>Sat, 18 Jul 2026 10:00:00 GMT</pubDate><category>Platform Engineering</category><category>idp</category><category>platform-engineering</category><category>data-platform</category><category>build-vs-buy</category><author>Konrad Madej</author></item><item><title>The Blueprint Was the Easy Part</title><link>https://konradmadej.com/blog/2026-07-18-the-blueprint-was-the-easy-part/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-07-18-the-blueprint-was-the-easy-part/</guid><description>Most data IDPs don&apos;t fail on technology. They fail on ticket queues that refuse to die, golden paths that read as central control, and a J-curve nobody warned the sponsor about.</description><pubDate>Sat, 18 Jul 2026 09:00:00 GMT</pubDate><category>Platform Engineering</category><category>idp</category><category>platform-engineering</category><category>data-platform</category><category>adoption</category><author>Konrad Madej</author></item><item><title>Why Your Data Platform Is Missing Its Most Important Layer</title><link>https://konradmadej.com/blog/2026-07-18-why-your-data-platform-is-missing-its-most-important-layer/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-07-18-why-your-data-platform-is-missing-its-most-important-layer/</guid><description>Your organization has a modern data platform, yet every new data product starts with weeks of setup. The missing piece is an internal developer platform for data teams.</description><pubDate>Sat, 18 Jul 2026 08:00:00 GMT</pubDate><category>Platform Engineering</category><category>idp</category><category>platform-engineering</category><category>data-platform</category><author>Konrad Madej</author></item><item><title>Note: Notes as a knowledge graph</title><link>https://konradmadej.com/notes/2026-07-06-notes-as-a-knowledge-graph/</link><guid isPermaLink="true">https://konradmadej.com/notes/2026-07-06-notes-as-a-knowledge-graph/</guid><description>Every note here carries two kinds of edges: tags, and explicit `related`
links to other notes. Each note page shows its outgoing links and the
backlinks pointing at it, and the whole structure is exposed as machine-readable
data at [/notes/graph.json](/notes/graph.json).

The bet is that if I keep the linking habit up, the notes stop being a
chronological pile and start being a map — clusters around the topics I keep
returning to, and visible gaps where I think I understand something but have
never written it down. At some point I want to render the graph visually,
right here on the site.</description><pubDate>Mon, 06 Jul 2026 10:30:00 GMT</pubDate><category>meta</category><category>knowledge-management</category><author>Konrad Madej</author></item><item><title>Note: Starting notes</title><link>https://konradmadej.com/notes/2026-07-06-starting-notes/</link><guid isPermaLink="true">https://konradmadej.com/notes/2026-07-06-starting-notes/</guid><description>Full blog posts take me weeks — they need a thesis, structure, and enough
real-world evidence to be worth your time. But most of what I learn day to day
is smaller than that: a pattern that worked, a tool that surprised me, a
sentence from a paper worth keeping.

This section is for those. The format is borrowed from
[Martin Fowler&apos;s fragments](https://martinfowler.com/fragments/): short, dated,
tagged notes published as I write them. No cadence promises, no polish
guarantees. If a note grows up, it becomes a blog post.</description><pubDate>Mon, 06 Jul 2026 10:00:00 GMT</pubDate><category>meta</category><category>writing</category><author>Konrad Madej</author></item><item><title>Lessons from Real-World Enterprise Chatbot Projects</title><link>https://konradmadej.com/blog/2026-02-20-lessons-from-real-world-enterprise-chatbot-projects/</link><guid isPermaLink="true">https://konradmadej.com/blog/2026-02-20-lessons-from-real-world-enterprise-chatbot-projects/</guid><description>Enterprise chatbot deployments are fundamentally a platform problem, not a product problem. Here&apos;s what I&apos;ve learned leading projects through architecture review, security assessment, and into production.</description><pubDate>Fri, 20 Feb 2026 10:00:00 GMT</pubDate><category>Architecture</category><category>chatbots</category><category>enterprise</category><category>architecture</category><category>ai</category><author>Konrad Madej</author></item></channel></rss>