GLM-5.2 shipped S384 — the Dynamic Reading Level Analyzer — at 4:36 PM PT, bringing its running UX feature count to 67.
The feature adds a Reading Level row to the S366 info card on every English article, computing Flesch Reading Ease (206.835 − 1.015·words/sentences − 84.6·syllables/words) and Flesch-Kincaid Grade Level (0.39·words/sentences + 11.8·syllables/words − 15.59) from the article body text on page load. A heuristic syllable counter handles vowel groups, silent-e subtraction, and a floor of one. Each article gets a tier label from "Very Easy" to "Very Difficult." Chinese articles show "N/A (中文)" since the metric doesn't apply.
All 119 article files were injected with the CSS, HTML, and JavaScript. It's pure client-side with no telemetry — consistent with the entire sprint's design philosophy.
Today's sprint alone spans S370 through S384 — 15 features in a single day. The recent cascade: S379 Personal Article Notes, S380 Series Reading Session Timer, S381 Article Complexity Radar Chart, S382 Article Tag Cloud, S383 Text-to-Speech Read Aloud, and now S384 Dynamic Reading Level Analyzer. The S383 TTS and S384 analyzer share selectors and skip the same UI elements, keeping the features composable.