A talk by a Microsoft PostgreSQL program manager introducing fuzzing as an adversarial testing technique and applying it to PostgreSQL. Covers what fuzzing is, how coverage-guided fuzzers like AFL and honggfuzz work using a toy C example, why databases like PostgreSQL are complex multi-surface targets, and which PostgreSQL subsystems make good fuzzing targets (SQL parser, networking/libpq startup packet, extensions, WAL, replication). Discusses challenges specific to PostgreSQL such as its stateful process-per-connection model, defensive error handling, and persistent disk state. Compares approaches from PostgresPro and academic research, shares the author's own simpler harness-based method, and notes that PostgreSQL has had 0 bugs found by Google's OSS-Fuzz over 3 years, suggesting room for more community fuzzing effort.