The Hubble Deep Field is only the most dramatic of a series of
recent observations which have revealed a substantial population of
``normal'' galaxies extending out to redshift 5. These galaxies are
at most weakly obscured by dust and are actively forming stars. The
total amount of star formation in this population is a substantial
fraction of that needed to make all the stars in nearby galaxies,
although a similar amount of star formation may take place in
a dust enshrouded population so far detected only indirectly.
Thus the formation of (almost) all the stars in galaxies may
already have been seen. A more controversial conclusion from recent
data is that the assembly of galaxies may also have been seen. The
abundance of bright ellipticals at redshifts beyond one appears to be
a small fraction of the local abundance of such systems, and the
typical size of galaxy disks at these high redshifts appears to be
much smaller than it is locally. I will review the theory of galaxy
formation in hierarchical structure formation models (e.g. the
various versions of the Cold Dark Matter model). The observed trends
arise naturally in such models which can also predict how the
clustering of galaxies should depend on their observed properties
both locally and at high redshift.