The "All" feed stuttered because every row appearing ran markSeen, which did a disk save(), an O(n) contains over a @Query'd SeenStory array, and triggered a full feed re-render on each insert. Now markSeen does an O(1) insert into a non-observed reference Set, no per-row save (the container autosaves), and SeenStory is no longer @Query'd by the view. cacheStories also does one fetch + an in-memory index instead of a query per story. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>