#[[substack]] Since starting to use LLMs, I write a lot [[more than]] I used to. I am not saying that I am writing many more papers than before, although I am a fair bit quicker at finishing papers than I used to be, I can't be sure that this is entirely LLMs as opposed to me getting better at this generally. What I actually mean is that days work for me consists of me either typing or speaking to an AI, or reading or listening to back to its responses. Typically there will be multiple LLM instances I'm jumping between. This is also not even including the quick AI and AI commands that I use through [Raycast](https://www.raycast.com/core-features/ai). Yeah, it's fair to say that my working life, the research part at least, is seeped in LLMs getting excited about delving into things. A significant portion of this [[Substack]] is going to be about me trying to articulate what exactly I'm doing when I'm in this frenzy of delving into things %%this is just a place holder, I don't like the way this joke is made here or above, it needs to be fixed or removed%%. Today I am going to talk about one of them. We can start by going back to what I find a lot also in the humanities more generally—there’s a strong suspicion that the LLM is doing the work, while the user sits and watches, or looks at short form videos. What I’m doing is taking a very literal approach to the phrase “prompt craft,” treating it as actual craft—as if I’m making an object.