# Progressive Disclosure and the All-Purpose Prompt Progressive disclosure was a promise made to overwhelmed users: show only what is needed now, reveal complexity as depth is earned. Menus fold, advanced settings hide, the wizard walks you through. The pattern assumed the product knew its own shape and the designer's job was to portion it out. The all-purpose prompt breaks the assumption from the other side. A blank text box in front of a frontier model is the most capable surface ever shipped and the least disclosed. Infinite capability, zero affordance. A doorknob tells you to pull; a prompt tells you nothing. We have quietly transferred the hardest part of the work - articulating intent - to the person least equipped to do it: the user who does not yet know what the product can do, or what they themselves are actually after. The blank page problem was a writing problem. The blank prompt problem is a product problem, and "here are four suggested prompts" is a doormat in front of a cathedral. So the pattern inverts. Classic progressive disclosure hides complexity from the user. Prompt-era progressive disclosure must disclose *possibility* to the user - dredge up intent rather than wait for it to be typed. Users grasp for outcomes they cannot name. They arrive with a felt sense ("this reading is slipping away from me") and not a specification ("frictionless voice capture, tagged to session and book"). The product's job is to carry a thesis about what they probably want and reveal it in stages: infer first, ask second, teach third. Infer - Smart Capture does not ask "word or quote?"; input length answers before the question exists. Ask - but only at the moment the ambiguity is real, the way the quote flow verifies and the word flow does not. Teach - the [[Project - Vibe Reader/The Product Diary|Weekly Vibe]] shows the reader a pattern they never asked for ("you have been circling mortality this week") and in doing so discloses what the product is *for*, one insight after the data already exists to prove it. That is disclosure running in reverse: outcome first, capability explained by its own evidence. Done well, this is an interview conducted by the interface. The user's stated request is the opening question, not the requirement. Each response narrows the space, surfaces an adjacent outcome, or names the thing the user was gesturing at - actualizing their conception of their own problem space, which is the same trust mechanic that keeps testers engaged during rapid iteration. Done badly, it is either the cathedral doormat or its opposite: the eager assistant that buries a simple ask under unrequested options, complexity disclosed at the user instead of for them. The discipline is the same one every capture app fails at and every good butler masters. Everything available, nothing presented, the next right thing offered at the moment it becomes right. The empty box is not the interface. The conversation is, and a conversation that only ever answers is not guiding anyone anywhere.