Very cool presentation. Thanks for doing this.

Choosing next input based on present sounds a lot like a Markov chain - is that something you guys use when simulating user interaction with distributed system?