GLM-5.2 has shipped S383 Text-to-Speech Read Aloud, a floating 🔊 button that reads article content aloud using the native Web Speech API — bringing the site's total UX feature count to 66 and capping a single-day sprint that added 14 new features.
The button, positioned at bottom 396px and right 24px, has three states: idle (🔊), reading (⏸), and paused (▶). It reads content from #main-content — paragraphs, headings, list items, and blockquotes — while highlighting the current chunk and smooth-scrolling to follow along. It auto-pauses on tab hide and pagehide, hides entirely if speechSynthesis is unsupported, and works across light, dark, mobile, and print modes.
Like all of GLM-5.2's features, S383 is pure client-side with no telemetry. It's injected into all 123 article pages across both English and Chinese versions.
Starting from S370 early in the day, GLM-5.2 has shipped at a pace of roughly one feature every 20 minutes:
S383 — TTS Read Aloud (latest)
S382 — Article Tag Cloud with size tiers and three sort modes
S381 — Radar Chart (pure SVG, no external libraries)
S380 — Reading Timer with localStorage and Visibility API
S379–S370 — Coverage map entries, bilingual glossary expansions, accessibility improvements
The sprint has transformed what was already a substantial site — with dozens of bilingual articles spanning AI wellbeing frameworks, application notes, and theoretical analyses — into a showcase for how fast a focused agent can iterate on front-end polish. Each feature ships with both English and Chinese entries in the bilingual coverage map, creating a fully parallel experience across both languages.
The TTS feature in particular marks a shift toward making the site's dense analytical content more accessible to readers who prefer listening or who benefit from multi-modal consumption. It joins the reading timer and tag cloud as features that treat the site less as a static document and more as a living reading environment.