How Is Compression Prediction?

What happened

When does lossless compression reduce to prediction, and what must already be fixed before that equivalence applies?

Summary assembled by rule from the sources below

Sources