Stellar spectroscopy is going through a revolution, changing manual labor by automated processing and artificial intelligence, and replacing human observers and single-target instruments by robots and highly multiplexed spectrographs. Yet, improvements at the fundamental physics level are happening far slower, and the most popular stellar atmosphere models employed to interpret stellar spectra are based on codes originally written in the 1970's. In this talk I will discuss recent advances and reflect on what's missing to realize actual progress in our understanding of the assembly and early evolution of the Milky Way from the study of the most primitive stars. To arrange a visit with the speaker during the visit, please contact their host: Hans Ludwig (LSW)