I will discuss recent and ongoing work focused on attempts to restore concordance amongst cosmological data sets, motivated by discrepancies between some measurements of the cosmic expansion rate (H_0) and the matter clustering amplitude (S_8). Particular attention will be paid to models invoking new physics in the high-redshift universe, including quasi-accelerating early dark energy models (and extensions thereof, featuring EDE-dark matter interactions) and generalized decaying particle scenarios. In particular, I will discuss constraints on these models derived using the latest CMB measurements from the Atacama Cosmology Telescope (ACT) and from the Planck satellite, amongst other data sets. I will present new constraints on canonical EDE models derived from previously-unconsidered late-universe data sets, which effectively eliminate the viability of this scenario to resolve the Hubble tension. Finally, I will highlight new, extremely precise neural-network-based emulators for Boltzmann codes that we have trained in recent months, which enable Stage-IV-level cosmological parameter inference to be performed in minutes on a laptop.