The market resolves to "yes" if they publish data/estimates for either task description or use case. I would call "summarization, translation, and code completion" examples of task descriptions. And "homework automation (cheating), medical or mental health consultations, generating marketing copy, testing the capabilities of the model, literature review, learning new topics/skills, and email automation" examples of gesturing at audience/purpose.

I'm imagining something of the shape of "Our model has been prompted X times. 3% of the prompts were to A, 5% of the prompts were to B...".

I would consider gestures of this in tweets, podcasts, talks as passable even if there isn't any official publication.

On the one hand, this is valuable market data. On the other hand, there might be PR advantages or pressures or transparency strategies.