<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Capstone on DATATWEETS</title><link>https://datatweets.com/courses/llm-evaluation/capstone/</link><description>Recent content in Capstone on DATATWEETS</description><generator>Hugo</generator><language>en</language><copyright>Copyright (c) 2026 Datatweets</copyright><lastBuildDate>Tue, 07 Jul 2026 09:00:00 +0200</lastBuildDate><atom:link href="https://datatweets.com/courses/llm-evaluation/capstone/index.xml" rel="self" type="application/rss+xml"/><item><title>Lesson 1 - Guided Project: An Eval &amp; Observability Harness</title><link>https://datatweets.com/courses/llm-evaluation/capstone/lesson-1-guided-project-eval-and-observability-harness/</link><pubDate>Tue, 07 Jul 2026 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/llm-evaluation/capstone/lesson-1-guided-project-eval-and-observability-harness/</guid><description>The course capstone brings all eight modules together into a single evaluation-and-observability harness for Docent. Run live on the Meridian docs, it scores a deterministic 0.778 and a judge 1.000 on a nine-case golden set, confirms retrieval recall@2 of 1.000, reports 2101 tokens and about a third of a cent per run with a p95 latency near 1834 ms, and passes a production gate to SHIP — every number produced by the running code and traced back to the module that taught it.</description></item></channel></rss>