Such initiatives are actually great for the modern AI community, I am sure we could use this kind of behaviors to precise the internal state of models and helps us understanding more on these aspects