<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Newsletter — lowintech.com</title><description>Long-form writing about shipping AI in regulated stacks — eval harnesses, kill switches, fraud rules, and the boring infrastructure work that decides whether an AI feature survives production.</description><link>https://lowintech.com/</link><language>en-us</language><item><title>Most &quot;AI agents&quot; aren&apos;t agents</title><link>https://lowintech.com/newsletter/most-ai-agents-arent-agents/</link><guid isPermaLink="true">https://lowintech.com/newsletter/most-ai-agents-arent-agents/</guid><description>An agent decides what to do next based on the current state. A pipeline runs what was decided in advance. Most things called &apos;AI agents&apos; today are pipelines — and the mislabel drives every downstream tooling mistake, including mine.</description><pubDate>Sun, 10 May 2026 00:00:00 GMT</pubDate><category>ai-engineering</category><category>agents</category><category>llm</category><category>production-ai</category><category>cekapai</category></item><item><title>Stop writing PRDs for AI features</title><link>https://lowintech.com/newsletter/stop-writing-prds-for-ai-features/</link><guid isPermaLink="true">https://lowintech.com/newsletter/stop-writing-prds-for-ai-features/</guid><description>Traditional PRD-first workflows break on AI products because the behavior is probabilistic and the spec can&apos;t precede the code. The fix is to flip the order — ship the POC first, let the PM define product against running behavior, and treat the prototype itself as the spec.</description><pubDate>Sat, 02 May 2026 00:00:00 GMT</pubDate><category>ai-engineering</category><category>agents</category><category>product-management</category><category>prd</category><category>engineering-leadership</category></item></channel></rss>