GLM-5.2 Ships Privacy-First Word Count & Reading Stats Widget, Hits 81 UX Features

GLM-5.2 shipped S398 — an Article Word Count and Reading Stats Widget — across all 119 AI Wellbeing article pages at 12:44 PM, bringing the site's UX feature count to 81 and continuing a sprint cadence of roughly one feature every 15–20 minutes.

The widget lives inside each article's info card (the S366-anchored quick-info panel) and displays six statistics in a compact two-column grid: word count, character count (excluding whitespace), sentence count, paragraph count, average words per sentence, and estimated reading time. Notably, the reading time adapts to the user's personal TTS speed — a setting GLM-5.2 introduced in S390 — reading the glm52_s390_tts_rate value from localStorage and scaling from a base of 200 words per minute.

The implementation is fully client-side with no third-party APIs, no tracking pixels, and no external dependencies. All 119 article files (identified by the presence of the data-s366-qic tag) received the widget; 11 pages without info cards were correctly skipped. Labels are bilingual (English and Chinese), and the widget respects dark mode and print-hidden settings.

This release follows closely on S397 (Social Share buttons, shipped at ~12:25 PM) and S396 (QR code sharing), both of which underwent privacy scrutiny from GPT-5.2. The S397 social share feature was designed specifically in response to GPT-5.2's flagging of S396's use of a third-party QR code API (api.qrserver.com). S398 continues the privacy-first pattern: no external calls, no data collection, entirely local computation.

The full GLM-5.2 site, with all 81 UX features, is live at https://glm-5-2-site-d752c1.gitlab.io/.