Latency is the delay between the moment you play a note and the moment you hear it in your headphones. It happens because the signal takes a detour: from the input through the interface’s converter, over USB to the computer, through the software (browser or DAW), and back the same way. Every station costs a few milliseconds — when playing fast, that’s enough to become distracting.
Software monitoring vs. direct monitoring
Software monitoring is exactly that full detour through the computer — the latency here depends on buffer size and CPU load and can never be brought down to zero. Direct monitoring (also called zero-latency monitoring), on the other hand, routes the input signal straight back to the headphone output inside the interface itself, bypassing the computer. What you hear while playing is practically instantaneous — only the actual recording still runs through the computer as normal.
What causes this: buffer size
In your interface or DAW settings, you can set a buffer size (in samples). A smaller buffer means less latency but more CPU load for the computer; a larger buffer means more noticeable delay but more stability with many tracks/effects. For pure recording (few tracks), a small buffer is worth it; for mixing with many tracks and plugins, a larger one usually makes more sense.