Close Enough: The Shortcuts Behind Systems at Scale

Large-scale systems constantly need to answer questions like “have I seen this before?” or “how many unique things are there?” — at a scale where exact answers are too slow or too expensive to compute. The trick these systems use is deceptively simple: trade a small, known error rate for massive gains in speed and memory.

This talk is a beginner-friendly introduction to probabilistic data structures, built around 3 of core problems they solve — membership testing, duplicate detection, cardinality estimation, frequency estimation, nearest-neighbor search, and random projections. Rather than starting from theory, we’ll start from something every developer already knows (the hash map) and build up from there, with plain-language analogies for each structure.