One of the outstanding achievements of cosmology is that the state of the universe when it was only a few seconds old seems to be well understood. The details have firmed up, and we can make confident predictions about primordial neutrinos, and He and D nucleosynthesis. This progress, spanning the last 30 years, owed a lot, on the theoretical side, to David Schramm and his Chicago colleagues. The way the universe cools, and eventually recombines, and the evolution of the (linear) perturbations that imprint angular structures on the microwave background, is also well understood. But this gratifying simplicity ends when primordial imhomogeneities and density contrasts evolve into the non-linear regime.
The Universe literally entered a dark age about 300,000 years after the big bang, when the primordial radiation cooled below 3000K and shifted into the infrared. Unless there were some photon input from (for instance) decaying particles, or string loops, darkness would have persisted until the first non-linearities developed into gravitationally-bound systems, whose internal evolution gave rise to stars, or perhaps to more massive bright objects.
Spectroscopy from the new generation of 8-10 metre telescopes now complements the sharp imaging of the Hubble Space Telescope (HST); these instruments are together elucidating the history of star formation, galaxies and clustering back, at least, to redshifts z = 5. Our knowledge of these eras is no longer restricted to `pathological' objects such as extreme AGNs - this is one of the outstanding astronomical advances of recent years. In addition, quasar spectra (the Lyman forest, etc) are now observable with much improved resolution and signal-to-noise; they offer probes of the clumping, temperature, and composition of diffuse gas on galactic (and smaller) scales over an equally large redshift range, rather as ice cores enable geophysicists to probe climatic history.
Detailed sky maps of the microwave background (CMB) temperature (and perhaps its polarization as well) will soon offer direct diagnostics of the initial fluctuations from which the present-day large-scale structure developed. Most of the photons in this background have travelled uninterruptedly since the recombination epoch at z = 1000, when the fluctuations were still in the linear regime. We may also, in the next few years, discover the nature of the dark matter; computer simulations of structure formation will not only include gravity, but will incorporate the gas dynamics and radiation of the baryonic component in a sophisticated way.
But these advances may still leave us, several years from now, uncertain about the quantitative details of the whole era from 106 to 109 years - the formation of the first stars, the first supernovae, the first heavy elements; and how and when the intergalactic medium was reionized. Even by the time Planck/Surveyor and the Next Generation Space Telescope (NGST) have been launched, we may still be unable to compute crucial things like the star formation efficiency, feedback from supernovae. etc - processes that `semi-analytic' models for galactic evolution now parametrise in a rather ad hoc way.
And CMB fluctuations will still be undiscernable on the very small angular scales that correspond to subgalactic structures, which, in any hierarchical (`bottom up') scenario would be the first non-linearities to develop. So the `dark age' is likely to remain a topic for lively controversy at least for the next decade.