FINDING #38 · UNIT ID 1321790489
leonickson1/Swiftlet
Swiftlet is a Swift and Metal runtime that runs large Qwen Mixture-of-Experts models locally on Apple devices by streaming expert weights from storage, enabling 35B and 80B models to run with low RAM, including on iPhone.
SURPRISE SCORE
0.00
Score Breakdown
SCORE = SURPRISE × ENGAGEMENT^0.7 × FRESHNESS × VISIBILITY × CONFIDENCE
SURPRISE = WINDOW STARS / DAYS / (AUDIENCE + FLOOR)
35% OF STARS IN ARCHIVE
Growth Telemetry
VELOCITY /D
24.00
ACCEL
-6.00
RETENTION
13.9%
PEAK 2026-08-05 · FORK-RETENTION 37.5% · 168 STARS / WINDOW
Author Audience
AUDIENCE
70
FOLLOWERS
8
OWNER ★
618
Engagement Signals
FORKS
23
ISSUE AUTH
0
PR AUTH
0
UNIQUE STARGAZERS 168 / 168 (DIVERSITY 1.00)
Why This Is A Finding
leonickson1/Swiftlet собрал 168 звёзд за окно, тогда как у автора всего 8 подписчиков — эффективная аудитория ≈ 70. Это даёт surprise-индекс 0.22 (звёзды относительно охвата автора, а не в абсолюте). Удержание форков 37.5% и 0 внешних контрибьюторов отделяют реальный инструмент от разовой вспышки. Акселерация отрицательная — внимание остывает после пика.
Related Findings
RANKS ABOVE 100% OF 9721 FINDINGS
METRICS IN CONTEXT
MEDIAN ACROSS ALL 9721 FINDINGS · Δ vs MEDIAN · PERCENTILE = SHARE RANKED BELOW
METRICVALUEMEDIANΔ MEDPERCENTILE
SCORE0.240.00+0.24ABOVE 100%
VELOCITY24.003.43+20.57ABOVE 91%
RETENTION13.9%32.1%-18.2 PPABOVE 25%
FORKS23130-107ABOVE 17%
SURPRISE0.220.01+0.21ABOVE 99%