I wanted this bot to be a conversation bot, so I would need some sort of AI engine. I wanted it to be a learning bot, so it could learn from what people have said before. I searched around, and it seems there aren’t many good learning AI engines out there for Linux. I eventually settled on MegaHAL.
I then obtained the logs for the conversation in #haskell over the past year, and used it to train the bot.
The result was sometimes hilarous, and often riddled with typos and nonsensical (considering the source, many would say that would be expected).
I started a page of MegaMonad quotes with some of the funnier exchanges.