Last weekend, I attended Darius Kazemi‘s Bot Summit (via IRC while watching the livestream).
I generally expect to feel out of place at Gatherings of Knowledgeable People, but everyone was so friendly and patient with me and my questions. I met a lot of really cool people there. I would have liked to keep going, but hangouts are limited to four hours.
When it ended, I felt inspired and thought I should try making a bot. My previous ones had all been either in Django or on AppEngine, but after Darius told me Heroku was free for single-thread apps, I really wanted to try it.
My first idea was to use Markov chains to generate tweets from the text of any tweet that contained “fuck that” or “fuck you” and end it with “fuck that and fuck you”. I thought it would end up at least a little bit funny, but my searching yielded not much that I’d actually want repeated ever, so I’m really glad that one didn’t pan out. Except for some rare, obviously sarcastic tweets, it wasn’t really something I would have been okay with signing my name to.
I don’t want an angry bot. I want a happy bot. A bot like hanging out with Mr. Rogers over one-too-many cups of warm sake. So I tried “Imagine” and received a lot of happy, silly, or deep tweets. It felt like a really nice base to work with.
I got some really surprisingly good, funny, or deep results to begin with. Except after not too long, I realized I needed to exclude some really popular things to make sure the material was diverse (and slightly less graphic—One Direction and Bieber fans have REALLY active imaginations).
“Imagine your daughter in the future. Now imagine how perfect this moment was”
“Imagine the plot twist. Now imagine a single, dramatic tear rolling down my cheek as I stare lovingly at the charger”
“Imagine a ridiculously good looking so-and-so. Now imagine me naked”
“Imagine having a relationship like Mila Kunis and JT does in Friends with benefi ts . Now imagine Dragons”
Haha! This last one was just so non-sequitur. It was perfect. But then the dragons didn’t stop.
I didn’t know the AMAs were happening, and I had never heard of the band Imagine Dragons.
So I thought I’d change the filter again, which should have been pretty easy to do. I pushed a new copy to Heroku.
More dragons somehow. Three publishes later… there are still dragons everywhere. What. WHAT. There’s no way I’m that bad with regular expressions.
Publishing the code to a new Heroku app took care of the pesky dragons. I was a little sad to block them entirely, but as soon as people stop talking about the AMAs, I would like to see some dragons flying around out there again.
It’s been running like that without any real problems since then, and some very nice people have said some very nice things. I wasn’t really expecting anyone but me to like it, honestly, so these comments came as an enormous surprise. It meant a lot to hear that yeah, some of my ideas are good sometimes. I remember thinking earlier today, “I really hadn’t planned on spending my lunchbreak happy-crying on Twitter, but I’ll definitely take it!”
I wasn’t sure if it was okay to feel so attached to my bot, so I felt guilty being so happy about how it had turned out. But it kind of makes sense, doesn’t it? My bot is a piece of my imagination, floating around on its own, going on and on about boy bands and fairy tales, and putting together things it finds in surprising ways and holding them up for me to put on the fridge.
So that’s the story of why the bot’s icon says Dragons. I have no good explanation for why I chose Penny, except that I was looking through some old G+ posts and found a picture I took while watching Inspector Gadget and thought it looked mindblown, as I hoped the people reading my bot’s tweets would be. I thought about changing it back once Penny’s dragon obsession stopped, but I left it because I like the idea of bots having personalities and obsessions, even if it is only because of my mistakes.