My observation from this is that the process by which humans assemble and summarize information, at least at a somewhat high level, is pretty darn close to the way that LLMs do it. I think that falls apart when you want to talk about deeper learning, drawing inferences and so on but if you are just trying to pull together an executive summary on XYZ, an LLM with some fact checking gets you pretty far IMHO.
Up until recently I was editor at an AI company, and parts of this read exactly like some of the outputs I'd get out of GPT-4/ChatGPT Plus.
Many of the linked references are too recent to be in the training corpus (Llama 2, for example), so unless there's some web-search component to this it looks like an LLM wrote the first draft, and a human went through to edit, add links, and populate with images, etc.