The deep surveys have finally established that the cosmic X-ray background in the 2-10 keV band is produced by the integrated emission of obscured and unobscured AGN, confirming to the first order the main prediction of population synthesis models. Now it is possible to put tighter constraints to the model parameter space and study in detail the AGN evolution down to luminosities of 1042erg s-1 at z ~ 1, checking if previous assumptions were correct. The AGN redshift distribution observed in the deep surveys seems to peak at z < 1, while standard synthesis models were expecting a peak at z = 1.3 - 1.5. Assuming that obscured AGN evolve very quickly up to z < 1 does solve the discrepancy in the redshift distribution but predicts too many obscured AGN among low-z sources. A new determination of the XLF of unobscured AGN in the regime not covered by previous shallower surveys is therefore needed, which has then to be implemented into the XRB synthesis models.
I gratefully acknowledge all the members of the Chandra Deep Field South collaboration.