In the recent five years observers have made great progress in mapping the cosmic nuclear accretion history. The luminosity function of type-1 QSOs has been mapped over ever larger ranges of the redshift-luminosity plane. We have reached reasonable consistency between different surveys and have constrained most of the UV-optical AGN luminosity density within the range of existing surveys. We have found no evidence of spectral evolution - AGN appear to be always the same kind of animal, irrespective of the epoch of observation.
The census of optically-obscured (type-2) AGN is starting to catch up, but there is still much work needed to understand their contribution at redshift z > 1. Whether we look at obscured or unobscured AGN, at X-ray or optical luminosities, we probably see a pattern of cosmic downsizing among AGN, where high-luminosity AGN were the first active objects in the early Universe, while lower-luminosity AGN lagged behind and only became abundant later.
At redshift z < 2 we see AGN of either type being hosted predominantly by elliptical galaxies with little morphological disturbance and some extra light from younger stars. They do not appear particularly disturbed and do not undergo simultaneous starbursts. In principle, AGN could be a late phase of major mergers where the nuclear activity only commences after the merger has relaxed into a regular elliptical, although this is not the only permitted interpretation.
On the whole, it is still unclear how to constrain the physical evolution of AGN and how to place them into the context of galaxy evolution. New estimates of black hole masses in the more distant Universe and observations of host galaxies at very high redshift might still harbor quite a few surprises for us.
I am grateful to P. Osmer for comments improving the manuscript. This work was supported by a PPARC Advanced Fellowship.