i'm trying to figure out what user context actually belongs inside AI apps. not the creepy “track everything” version. more like: writing style, preferred tools, current projects, interests, maybe a limited summary from chatgpt history if the user approves it. the hard part is that too little context makes the app generic, but too much context feels invasive fast. if an OpenAI-powered app asked for specific consented context, what would feel reasonable vs too much? submitted by &...