We use high-resolution simulations to show that the current standard paradigm for the growth of structure in the Universe predicts the formation of a galaxy like our own to differ substantially from the classic ELS and Searle/Zinn pictures. On scales larger than the Local Group, the earliest star formation was extremely inhomogeneous, suggesting that the high redshift intergalactic medium should have large-scale abundance variations. The very oldest stars should be found today in the central regions of rich galaxy clusters. In the Milky Way's bulge and stellar halo little correlation is expected between age and metallicity. Some of the lowest metallicity stars may be relatively young. Many of the oldest stars may have high metallicity. Spheroid stars were formed before and during halo assembly, the oldest now lying preferentially at small radii, while low metallicity stars lie preferentially at large radii. The bulk of the Milky Way's stellar spheroid came from a small number of progenitors. It should show little spatial structure in the inner 5 to 15 kpc, but consist at each point of a superposition of hundreds of `cold' streams. Such streams have been detected in the Solar neighborhood.

9 pages, 3 figures. To appear in the proceedings of the 1999 MPA/ESO workshop "The First Stars" (eds A. Weiss, T. Abel and V. Hill)