Three months ago I started saving every twitter thread that includes an arXiv link and have been playing with heuristics to curate the most interesting conversations. Sample output from a couple weeks worth of papers published in mid-August