<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Yoseph Eikelman · Notes</title><description>I embed in a business, find the operational bottleneck, and ship production AI that people use every day.</description><link>https://yussieik.com/</link><item><title>A deploy cost us a 58-minute therapy session. Now an interruption costs one stage, not the session.</title><link>https://yussieik.com/notes/deploy-lost-recording/</link><guid isPermaLink="true">https://yussieik.com/notes/deploy-lost-recording/</guid><description>How ClinicLog turns a recorded session into a transcript in three saved stages, so a deploy, a shutdown or a killed worker loses at most the stage it lands in.</description><pubDate>Sun, 01 Nov 2026 00:00:00 GMT</pubDate><category>production</category><category>reliability</category><category>workers</category><category>healthcare</category></item><item><title>An LLM checks my production health four times a day. It gets counts, not log lines.</title><link>https://yussieik.com/notes/llm-health-gate/</link><guid isPermaLink="true">https://yussieik.com/notes/llm-health-gate/</guid><description>How ClinicLog, a medical SaaS, lets a model judge server health from counts while raw log lines stay on its own servers, and why the model never gets the last word on a critical.</description><pubDate>Sun, 01 Nov 2026 00:00:00 GMT</pubDate><category>production</category><category>llm</category><category>monitoring</category><category>privacy</category></item></channel></rss>