how and why @nerdgarbagebot works

As usual, I’ve made a few new Twitter bots since my last update, but I want to talk a little bit about how a particular one– @nerdgarbagebot— works.

a screenshot of @nerdgarbagepitches

I’ve been talking a lot on Twitter about how simple bots can be to set up. I use Cheap Bots Done Quick and the Tracery visual editor for mine, so I hardly have to look at code when I’m writing them– my screen actually looks like this:

a screenshot of the tracery visual editor

Both of these are free tools that you can use to make your own bot; all you need to do is register a new Twitter account to get started.

The results of nerdgarbagebot are pretty varied and rich, but the code is really simple– it only actually has 13 variables in it:

  • Work titles (“Jurassic Park”, “The Sims”)
  • Creators (“George Lucas”, “Nintendo”)
  • Elements (“mermaids”, “a plot”, “interpersonal drama”)
  • Comparative adjectives (“grittier”, “with swears”, etc)
  • Formats/mediums (“tv show”, “installation art piece”, “webcomic”)
  • Genres (“gothic”, “cyberpunk”)
  • Settings (“on the high seas”, “in a modern high school”)
  • Story types (“coming of age story”, “mystery”)
  • Characters/People (“Yoda”, “Joseph Gordon-Levitt”)
  • Character roles (“mentor”, “president”)
  • Audiences (“tweens”, “atheists”)
  • “Imagine this” intro phrases (“Imagine”, “I need you to picture”)
  • “Fund this” intro phrases (“Crowdfund this”, “Please support my”)

For each tweet, Tracery at random chooses from a list of formulas, which are written like this:

  • It’s a #genre# version of #titles#, with #element#.
  • #imagine# a #format# version of #titles#.
  • #imagine# a #genre# #format# version of #titles#, but with #character# as the #role#.

I’m increasingly realizing that the reason bots like this work really well– compared to ones like @BuzzFiendNews, which I am still struggling with improving– is that even though the format is simple, every phrase is something that the bot’s audience brings their own baggage and associations to, so the tweets tend to be more unique in and of themselves.

BuzzFiend is a lot harder to add variation to, because even though each tweet has different words in it, every tweet using a specific formula tends to be similar to others. A joke about eating humans tends to be pretty similar to other jokes about eating humans, whether those humans happen to be sidekicks or princes.

Thinkpiecebot, nerdgarbagebot and some of my other bots, especially @likeuberbut, manage to capitalize on people’s existing ideas. Thinkpiecebot’s funniness in particular comes from the unexpected combinations that it produces being put into the recognizable headline format, but doing that ended up being complex– I have over 50 formulas in it and nearly as many variables, and I’m constantly updating it so that it keeps up with the zeitgeist.

Nerdgarbagebot is brand new, and I’ll probably have to continue updating it to keep things current, but I’m having a lot of fun with it– I hope you like the results too!

Leave a Reply