series/why-randomness-keeps-sneaking-into-software

Why Randomness Keeps Sneaking Into Software

A random index is a tiny contract: say exactly what can happen and with what probability. I follow that contract through weighted choices, streams, PostgreSQL pages, approximate membership, search heuristics, latent variables, and LLM decoding. Each stop asks the same awkward questions. Which distribution did we actually implement? Which assumptions make it honest? What does exactness cost?

0 parts · complete