What does the UI teach the user about how the underlying system works? What mental model do users form around the system, and how accurate is it? What are the failure modes, and how risky are those?

What affordances do ai products have? They have been designed to be used a certain way. Or they may lack particular affordances (like a chat interface). So perhaps its affordances are unintentional. What kind of behaviors and outcomes does this result in?

Followed! I regularly visited scribbling.net back in the day :)

I highly recommend both Bike and Taskpaper - the sense of deliberateness in the design decisions in both apps is just 🤌

There is something about how the native feeling of Mac-assed apps just doesn't compare to anything else. In gaming, it's called "game feel". Both Jesse's apps have it.

When you read LLM output, there is no singular author. The words are a soup of training data arranged in a grammatically sensible order. There is no intent behind the words. There is no revelation of the author, because there is no author.

Adam, I’m sorry to hear this. It sucks.