An interactive explainer

A Bloom filter never stores a single name, and can still rule one out with certainty.

Ten million banned emails cost about 300 MB to keep in memory. There is a structure that answers the same question in 12 MB and stores no name at all. The price is strange: it can tell you "definitely not on the list", it can never tell you "definitely yes", and it gets more confidently wrong the fuller you let it get.

Bloom filters · Probabilistic data structures · False positives

Databases & Search · Intermediate · 10 min · The list that only says no — an interactive explainer on ViLearn, with a simulation and a knowledge check. All explainers.

Related: How a database finds one row in a million · Why two correct transfers can lose money · How Elasticsearch finds your needle in a billion documents