HTML entities: decode > < & etc. on sliced segments only
(entity offsets index the stored string; decoding before slicing shifts them)
Image lightbox: click any tweet/thread/article image to open full-screen
viewer; cmd+click follows <a> to open in new tab; arrow-key + ‹ › nav;
Escape closes; 1/N counter; ↗ open-in-new-tab link
Multi-image grid: 2 photos → side-by-side (180px rows); 3 → left spans
both rows; 4 → 2×2 (140px rows); single image unchanged
Empty media grid ghost: build photos/videoItems arrays first, render
.mediaGrid div only when at least one item resolved (advisory: never gate
on raw media.length when map items can all return null)
QT indicator: ↻ QT badge on tweet.is_quote_status === true entries
Modal shrinks for short content: height: 88vh → max-height: 88vh;
.preview-modal-body gets max-height: calc(88vh - 52px) so long threads
still scroll (advisory: don't rely on flex:1 once parent has no fixed height)
Video scrollbar leak: .preview-modal-body overflow: auto → hidden; each
child (tweet-wrap, video-wrap, iframe) manages its own scroll surface
Styled thin scrollbar on .preview-tweet-wrap (matches workspace rail)
ArticleRenderer: cover image and body images are lightbox-clickable;
opts thread through renderBlocksJSX → renderBlockJSX → renderAtomicJSX
