Recently on a slow Friday we trained Markov chains with our Slack history. We generated a separate chain for each Slack member. It culminated with a Slack channel emulator generating full chat logs of what members would typically say.

The programming in Python was suprisingly quick and easy, but we ended up wasting more time then expected by amusing ourselves with the results.

So if anyone’s thinking to try out Markov chains for text, I recommend taking Slack logs for training.