One of the most challenging problems in computational galaxy formation is modeling distant heating and ionization by locally produced radiation. Most Radiative Transfer (RT) techniques are very computationally expensive and limit users to poor resolution or post-processing thus decoupling the radiation from the dynamics of the simulation. We present a new efficient method for RT implemented in the SPH code GASOLINE aimed at full cosmological simulations. The method is tree-based (similar to a gravity solver) scaling as N$_sinklogN_source$ in the optically thin case and as N$_sinklogN_sourcelogN_tot$ in the optically thick case. Applications range from the reionization of the Universe to H$_2$ formation and destruction. First applications focus on FUV and EUV emission from Milky Way-type galaxies and how these affect satellites galaxies.