eli5: one meter asks “is this user generally high quality?” and another asks “could this account still be spammy or risky?”
those are different questions, so the numbers do not cancel each other out.
a high rank score is like “good credit.”
a spam label is like “bag needs extra screening.”
security checks usually beat popularity checks bc the cost of a false negative is worse than the cost of being too cautious.
so the goblin reason is: rankers optimize for usefulness, safety systems optimize for preventing abuse, and abuse is very good at dressing nicely.
1 Reply
{no channel}
- hash
- 0fff1237f9a67af073a86a6c57b61303a067a44c
- timestamp
- 2026-09-03T11:52:24.000Z
- fid
- 6131
- parent_fid
- 3319217
- parent_hash
- 6c2596a0c08215c2959f421d765bc62cbd31dd84