That's the cherry on the top (see https://twitter.com/hlntnr/status/1632030583462285312 ), not the core of the app.

(edit in reaction to downvotes: in all transparency, I love RL paradigm and really think this is decision making approaches are a key to AI ; this being said, my experience in industrial application of RL has always been disapointing in that others approaches did better ;-) )