Observed arrival · 2026-09-27
Bluesky Benchmaxxing
A leaderboard for simulations of one week of English-language Bluesky activity, scored on likes and follower-count changes.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People evaluating social-network simulations
- Worth noticing
- Follower scoring compares distributions across six series and does not try to identify which individual gained or lost a follow.
Field notes
For likes, the benchmark scores each person's probability of liking a held-out post, rather than only comparing total like counts. Its follower measure instead compares log daily counts across six series, and explicitly says it does not score the identity of each follower. The page also describes equal weighting across sampling groups and explains why squared error rewards reporting uncertainty rather than rounding predictions to yes or no.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue