The cosmic microwave background (CMB) contains a wealth of information about the early and the late universe. In this talk, I will focus on the search of primordial gravitational waves. Specifically, I will talk about “delensing” — constraining the lensing component in the CMB B-mode maps that we might reduce the soon-to-be major uncertainty of the r measurement in the BICEP/Keck experiments. For next-generation CMB experiments, we will need higher signal-to-noise estimates of the lensing potential beyond the commonly-used quadratic estimator approach today for delensing. I will discuss a method to extract the lensing field using a convolutional neural network (1810.01483), that approaches maximum-likelihood lensing estimates in a broad range of angular scales. If I have time, I will discuss the “H_0 problem”, where the inferred expansion rate of the universe from the CMB (and a couple other probes) are significantly lower than direct measurements. I will look at it from the sound horizon perspective (1811.00537) and show how data from the South Pole telescope can probe potential new physics.