The most simple definition of "legacy code" is code that was written by someone else before you. However, this term has come to mean SO much more. Legacy code is essentially untested code, and code that has not yet been tested is basically bad code. Michael writes (in a wonderfully conversational way) about finding productive ways to work with legacy code (without completely wanting to cry and throw your computer at the wall).
This, together with Fowler's Refactoring is a must read for anyone getting into the field. We encounter legacy code all the time, and especially if you're coming in as a Junior, since they're usually charged with maintaining it. Great posts @katesegrin !
My Octocat by Github
Superwall