"They just completely ignore you?"

"Yeah. But to be fair that was after, like, over five minutes of me repeating the same question in the same way. First few times were fine, then they started getting annoyed, then they were annoyed and stopped answering, then they just started ignoring me altogether. Leaving and coming back ten minutes later got them to respond again—though they were still kinda annoyed—but I didn't explore that particular quirk much more deeply than that."

"That seems weirdly - specific, I don't know what I was expecting though."

"You can borrow my notes on stuff to try if you want. I've found that the best way to keep them happy is cycle through like four different things and add in a fifth every few minutes, if you hit the specific notes each of them likes you'll get a new friend pretty quickly."

"I'd like to see the notes."

He can text her a link to the document he's been maintaining.

"I didn't really organise it for consumption by others but hopefully it's still readable enough."

"Understandable," she snorts, and she opens it up.

The first page is dedicated to his up-to-date general formula for how much the "relationship number" changes depending on interactions. There seem to be two, actually, with the second one labelled as "romance/sex" and a comment saying "NTS: explore this when it feels less ick", but most of the work was done on the first number.

The number has a floor and a ceiling, which he arbitrarily set to -100 and 100 respectively, and the delta per interaction depends on the current number, "interaction type", a modifier for the subject's interest in the topic, a "mood" modifier that affects the effect of the interaction type, and one final modifier that he called "recent interaction" but which has another comment saying "NTS: needs more refinement, see section D".

The next several pages include an extensive categorisation of interaction types with commented examples and some sections on whole bits of interaction transcribed after the fact. Each type has a base value and a modifier value for whether the subject is neutral, mildly positive/negative, or very positive/negative on the topic.

After that he elaborates on the other modifiers, which seem to be multiplicative and which have multiple notes about needing more research on this specific thing or that.

And finally the whole second half of the document is taken by speculation on the meaning of each value band and predictions that can be made based on them. There's a threshold for when people start spontaneously calling you, and then various bands like "if the value is between 20 and 50 the subject stops reacting negatively to mentions of topics they find mildly negative, if it's above 50 they also reduce their reactions to very negative topics to the same they used to have to mildly negative ones". He's registered observations that led to each of those estimates and predictions, as well as observations that conflict with them, with some speculation on reasons and assumptions that may have been broken.

Finally, at the end, he has an extensive list of things to try at some point, with particular emphasis on ideas that can falsify specific assumptions and predictions by the model.

The whole document is clearly meant to be used by him, with notes to self being very common and paragraphs not caring so much about broad readability or non-redundancy. Still, it's pretty well-organised, with headers and subheaders that can be clicked on from the document's table of contents and comments labelled as things like "speculation" or "evidence" or "unconfirmed prediction" or "test to formalise" or similar.

"Wow, this is... rigorous."

"Not needing to sleep and having an intensely menial job I've refused to become an NPC for has left me with... a ton of free time."

"I take it you can't be a professional wizard."

"No one pays just for that, as far as I can tell. It does help with, you know, making ordinary life easier, like the not sleeping, but also it takes a very long time to develop, and just like chess it seems I progress more slowly than NPCs. But the upside is that the standard meditation for magic practice is simple enough to do I can usually do it at the same time as other things, like exercising or even some work stuff that doesn't require a ton of thinking."

"What's your job?"

"Oh I never did say, did I. I'm an assistant manager at Dewey, Cheatem & Howe Incorporated—it's an investment firm—and in practice what I do is be a glorified secretary and take on the bits of management my boss hates the most."

"The making friends thing doesn't have a working variant where you get promoted?"

"It does! I'll probably take my boss's position as soon as he retires, which should happen next week at the latest. But the whole 'life on fast-forward' thing applies to this, too, other people get promoted much more quickly than I do. Of course, maybe I just suck and shouldn't blame this on the NPC thing."

"Do they seem qualified for the work they're promoted to?"

Vague handwave. "In a certain sense? Just like they seem to learn how to choose the right chess moves when they play more chess they seem to spontaneously acquire better skills at, like, writing reports. I think 'being a person' by itself sort of beats their advantages at sounding persuasive or charismatic to each other, though, which might've helped me some."

"Huh. I'll have to try some of the same things you did, I guess."

"I'm not sure which kinds of things in journalism look more like chess and which look more like being able to model people but it sounds naively like the latter kind will dominate."

"Well, that assumes that they like the same things about news articles that I do, like 'not sounding like an NPC wrote them'."

"Not necessarily, even if you never use logical connectors between your sentences I think having access to them would make you better at writing stuff that they appreciate."

"I guess I'll find out. Right now I'm still doing a lot of fetching coffee."

"Oh right that thing where most jobs seem to for some reason involve a lot of that. When I started out I was mostly stuck in a room organising mail. Even though no one uses mail. It was mostly orders off online stores that people had sent to the office instead of their houses."

"No one uses mail?"