We propose a dedicated, ultra-deep spectroscopic survey in the near infrared (NIR), that will target a mass-limited sample of galaxies during two of the most fundamental epochs of cosmic evolution: the formation of the first galaxies (at z ≳ 6; cosmic dawn), and at the peak of galaxy formation activity (at redshift z∼ 1–3; cosmic noon). By way of NIR observations (λ= 0.8–2μ m), it is possible to study the UV Lyman-α region in the former, and the optical rest-frame in the latter, allowing us to extract fundamental observables such as gas and stellar kinematics, chemical abundances, and ages, providing a unique legacy database covering these two crucial stages of cosmic evolution. The need to work in the NIR at extremely low flux levels makes a ground-based approach unfeasible due to atmospheric emission and absorption. Only with the largest facilities of the future (e.g. ELT) will be possible to observe a reduced set of targets, comprising at most of order thousands of galaxies. Likewise, from space, the small field of view of JWST and its use as a general purpose facility will yield a rather small set of high quality NIR spectra of distant galaxies (in the thousands, at best). Our project (codename Chronos) aims to produce ∼ 1 million high quality spectra, with a high S/N in the continuum, where information about the underlying stellar populations is encoded. The main science drivers are: The connection between the star formation history and the mass assembly history.The role of AGN and supernova feedback in shaping the formation histories of galaxies, with a quantitative estimate of quenching timescales.The formation of the first galaxies.The source of reionization.Evolution of the metallicity-mass relation, including [α/Fe] and individual abundances.Precision cosmology through detailed studies of the “baryon physics” of galaxy formation, probing the power spectrum over scales k∼ 1 Mpc− 1.