Although this article is specific to artificial intelligence (AI) papers, it has a broader application, and I find many of the principles apply to papers in other domains, including online learning. A paper is not like a novel you read from beginning to end. Different types of papers are read differently. For example, I will read a formal academic paper starting with the abstract, then decide to look at the literature review (if the area is new to me) or the discussion (if it's not). If the discussion offers interesting results, I'll look at the methodology. For a stream-of-consciousness blog post, by contrast, I'll often read from the bottom up, so I get the author's main point first, then work back through the chain of reasoning that brought them to it. The main point here is: know what you're looking for, and have a strategy for finding it in the paper (if it's there). Image: Flame-In-Nerf.