Where did March go? OFFLFIRSOCH post
This year is the first year I've actually tried to participate in OFFLFIRSOCH, and I think I overestimated the time I'd have to work on it. It wasn't an unusually busy month, though I did have one offspring's birthday and a weekend mini-break with my girlfriend, so I did have some of my weekends AFK. The idea I had didn't seem like too ambitious a project, but around the 23rd, I found myself with only the basics of the project done. I've tried to make more time for it the last few days, but note that it is also Tax Season, so I've had that hanging over me [1], and I didn't quite finish. I'm releasing it anyway, even though it's not quite in a usable state. I'll finish it up over the next few weeks.
tarot.el, a Tarot assistant for GNU Emacs
My goal was an offline Tarot assistant. I am learning Tarot reading, and for card meanings, I tend to find myself checking websites for the meanings of cards, especially in the Minor Arcana, which has more cards and which I don't know as well. Of course I have books, but they are at home, and ebooks, but they are on my personal laptop. I'm likely lacking both of those whenever I'm not at home. However, I'm probably not lacking Emacs, since I use it at work, and can sort-of use it on my phone. It will also be quicker than these other methods when I'm at home, too.
The holdups on the development, besides the other things going on in my life, have been two. First, although I'm pretty proficient with Emacs Lisp, I'm not actually all that familiar with a lot of Emacs's display features, like images, faces, and text properties. So I've had to learn how to do those for the output, which is still not quite finished. The other is that there's just a lot of data entry. I am pulling the descriptions of cards and their normal and revered interpretations from Arthur Edward Waite's _Pictoral Key to the Tarot_ because it matches the most commonly used Tarot deck, and because it is in the public domain. So to define a card, I find the card in a text file of the book, write out the (make-tarot-card) definition in my code file, and fill in the card's slots with things I copy and paste from the book. It's a pretty mechanical process, but not so mechanical as to be easy to automate, which of course makes it not the first leisure time activity I want to turn to. All of this means that as of today, I have finished the data structures and algorithms (the fun stuff!), have a proof of concept for displaying a single card, and have the Major Arcana and three suits of the Minor Arcana entered in. I guess there's about an evening's work left to go on this.
Anyway, I've made the current state of it public on my Gitea instance.
tarot.el git repository
The looming spectre of AI slop
As a point of pride, I never use LLM-based tools in any of my hobby craft. But I have this nagging feeling that I could have finished this project this month with it? Not so much for the coding, but for the data entry. I expect an LLM probably could follow my examples for a few cards, and pull approximately the right text from the text file to fill them out. Heck, if I were even less scrupulous, I wouldn't even have to give it the text file; just let it fill out the values with the average of a bunch of Tarot web pages in its training data. So this is frustrating and discouraging to me on a number of levels. Usually I'm not very sympathetic to the idea that LLMs help with boilerplate in code, because good programming languages let you eliminate boilerplate. But this data entry has me vexed.
[1] For my non-USAian readers, the US has an unusual income tax season. Every year, each resident with an income must fill out their own tax forms to calculate the tax they owe, even though all the inputs of this calculation are known to the federal government, and they can and do calculate it themselves so they can punish you if you get it wrong. In the time before software ate the world, you could fill out paper forms, or pay someone to do it for you, if you brought in all the forms with the inputs. If you did it yourself, it didn't cost anything extra. In the early web era, you had the option of paying for a program either on your own computer or a website to submit it for you, or filling out the paper forms, or, for a limited time and for some people, to submit it for free through a federal website. Now in the enshittification era, you cannot file your own taxes on paper or through a federal website. You must pay a licensed third party, either in person or through a website, to submit it for you. That is, you must pay someone to send the federal government data they already have and to perform calculations they will also perform. This is considered the best possible system.
OFFLFIRSOCH 2026 Announcement