On @MythologyBot

I was recently asked some questions about my most-followed bot child, and decided to take the opportunity to craft a blog post mostly out of my responses to those.


spun off of a much more complicated idea to bring together personality testing and tarot/astrology to generate tailored fortunes/narratives based on responses to prompts. When I realized that was a lot of work to just make the point “Myers-Briggs is astrology for people who want to pick their sign” it got mothballed. Until…

BotSummit 2014

and specifically, thrice giving a talk about corpus selection and manipulation. I’d come across Stith Thompson’s Motif-index of folk-stories during research for the larger “I am Thou” project, and when thrice mentioned preserving the structure of the text and importance of “voice” I realized the motifs themselves are enough, if I just massage them a little to imbue a voice.

Inspirations and Prior Art

@MicroSFF (not a bot) was created in April 2013, right after Twitter added support for line breaks, so you could separate text vertically, and it was already very popular by April 2014. Line breaks were also a key feature in a lot of so-called “Weird Twitter” joke formats around that time. So putting line breaks in the motifs gives aesthetic variation and variable dramatic/comedic pacing, in those same modes. The format follows in the path proven by those earlier inspirations.

Although I didn’t find out until 2015, someone had already made a bot from the Thompson index 21 months before Mythology Bot: @StithThompson. Comparing the two side-by-side shows the impact stylistic choices I made based on thrice’s talk had, and they’re pretty significant! Its creator(s) apparently pivoted to a more ambitious concept which they never got off the ground, so there’s a teachable moment as well.

Selecting and Appreciating the Corpus

For immediate appeal, there were so many motifs that were under 140 characters after attribution/references were trimmed from the end, it was a source that could run for years without getting repetitive. Allison Parrish’s @everyword was just reaching the “w”s at this point, after nearly 7 years of tweeting, so longevity was on my mind.

My partner was a Classics major and I’m a bit of a secret Jungian, so I was familiar with the ideal of cataloging a finite taxonomy of stories. I’d read Neumann and Campbell and others, but hadn’t seen the Thompson index specifically until I was searching for corpora to make a bot. In hindsight it turns out my partner had a Thompson book (The Folktale) on our bookshelf for most of our time together but I’d never picked it up!

Some of my reliable favorites are the “clever” bits, the ironies, and the ones about fools and “numskulls”. I’m also partial to motifs that have such specificity they border on the fetishistic. And I have a soft spot for motifs built on returning from the dead—the thing we most universally need to go to stories (or faith) to find.


As far as my goals, my favorite endorsement for Mythology Bot might be this one:

I like to see something that isn’t determined by the news of the day, or what my friends are talking about. “Dumb” bots are an alien presence, and unlike the “smart” content delivery systems everyone is rushing to develop, embrace that fact. When that “dumb” thing can cut through the dominant conversation and grab you, that’s when it’s doing its job.

I’ve seen folks recommend/value it as inspiration for games, or NaNoWriMo, or tabletop game campaigns. I’ve seen folks get that identification kick: “it me” or “@(someone else) this is you.” I’ve seen folks work through or rail against current events through a motif coincidentally tweeted at that time. I’ve seen folks sharing an instance of the motif in response to the bot. All kinds of reactions! And each one makes me happy.

It has more followers than any of my other bots, but even proportionally it receives way more “engagement” than anything I’ve made. People like to talk at/about/around it, even knowing it’s a bot, which is the best possible outcome.


Mythology Bot tweets every 3 hours, at 44 minutes past the hour. With the number of motifs available for tweeting, this means it can tweet for about 15 years before running out. The selection is pseudo-random and non-destructive, though, so it can repeat itself, and already has a few times.

I like to follow (and make) bots that put out “a handful” of tweets daily. Bots you won’t forget are around, but you also don’t wish they’d shut up. It’s 44 minutes past the hour because I have other scheduled tasks on the same machine and since it runs quickly it would safely complete before an existing ***:45 task would start to run.

The scraper I built to pull the motifs down is lost to time, but the text file of possible motifs and the shell script that tweets them out (and a change history of me removing offensive motifs) is on GitHub.