The For You feed was chronological, hashtag-interest-EXCLUSIVE, and ignored
follows entirely (docs/feed-algo-audit.md). Now it is inclusive + ranked:
- createForYouScorerStage scores each video: follows-boost (author you follow)
+ interest/watch-topic affinity (cosine vs your interests+watch-history vector)
+ freshness - disinterest penalty. Reuses Explore's vector/similarity helpers.
- createForYouScoreSorter ranks by that score (muted last, recency tiebreak).
- For You pipeline: interests now BOOST not exclude (enforceInterests:false);
scorer added; chronological sorter replaced. Never empty (falls back to
recency when you have no follows/interests/history).
- buildForYouFeedRuntime now feeds subscriptionAuthors + watchHistoryTagCounts.
Watched/blocked/muted are still removed (suppression/blacklist/moderation
stages unchanged). Cheat-resistant tests (older boosted items only win via the
boost). Grid renders end-to-end (~1.3s). Weight tuning is a follow-up.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
- Introduced `NORMALIZED_ARRAY_MARKER` in `js/feedEngine/utils.js` to tag arrays that are already normalized.
- Updated `normalizeItems` in `js/feedEngine/engine.js` to skip O(N) verification if the input array is marked.
- Updated built-in stages and sorters to mark their output arrays using `markAsNormalized` when they safely construct new arrays from normalized inputs.
- This avoids redundant O(N) scans after each stage, improving pipeline performance by ~10% end-to-end and eliminating overhead for trusted stages.
Co-authored-by: PR0M3TH3AN <[email protected]>
Hoisted the timestamp hook resolution out of the inner loop in `createChronologicalSorter`.
This avoids allocating an array of candidate functions for every comparison during sorting (O(N log N)).
Benchmark shows a ~27% speedup (38ms -> 27ms for 100k items).
Co-authored-by: PR0M3TH3AN <[email protected]>
- Implements an Inverted Index to reduce redundant similarity checks.
- Replaces O(N) array splice with O(1) Map deletion for candidate tracking.
- Reduces runtime from ~3.5s to ~0.9s for 5000 items (3.9x speedup).
Co-authored-by: PR0M3TH3AN <[email protected]>
