Takeaway from roughly skimming the article: the single pixel detector is a plasmonic nanoantenna, which seems to be tunable. This is placed behind a set of optical diffraction gratings that have been optimized to the task using machine learning and essentially transform the spatially encoded information in the input into a spectral encoding. Both the gratings to encode the information and the decoder have been designed using machine learning. There is no dynamic illumination and the imaged object is static in front of the optics. All of this was done in the THz range.