RDB is a photograph, AOF is a diary: a crash costs what happened after the last entry.

An RDB file is a point-in-time snapshot of the whole dataset, so everything written after the latest snapshot disappears with a crash. AOF appends each write command and replays it on restart, and with everysec only about a second is at risk. Many setups combine both, using AOF for freshness and RDB for backups.

Read it once, then a question tests whether the trick stuck.