The question gets better every day you write.
Alice asks one question per day. For the first thirty days, the questions come from a curated seed set. After that, each question is generated from everything you have written, how you wrote it, and what you have never been asked. The longer you write, the more precisely the question finds you.
Days 1-30: the seed questions
Before Alice knows anything about you, the questions are general-purpose. They are designed to be worth answering regardless of who you are, and to produce writing that teaches the system about your patterns.
What is something you believe now that you did not believe five years ago?
Describe a decision you made recently where you are not sure you chose correctly.
What would you do differently if no one would ever know?
These questions are good. They are not personal. They could be asked of anyone. Alice is listening, but it has not yet learned enough to speak to you specifically.
What accumulates underneath
You never see any of this. But with each entry, Alice is building a model of your writing:
What you write about. Which themes recur. Which topics you return to. Which you avoid. What you said three months ago that connects to what you said yesterday.
How you write. Where your pauses fall. How your production speed changes with different kinds of questions. Whether you revise heavily or commit on first pass. Your keystroke rhythm is as individual as your handwriting.
Not what you say you feel. What your word choices and sentence structures reveal about how you process. Hedging language. Self-referential density. Cognitive processing words. These patterns are visible in the text whether or not you are aware of them.
What you have never been asked. What kinds of questions produce short, guarded responses. Where the blind spots are. A good question sometimes goes exactly where you have not been looking.
The calibration data tells Alice about your current state: sleep, stress, physical condition, daily rhythm. The question adjusts. A difficult abstract question on a morning when you slept badly is not depth. It is cruelty.
After day 30: the questions find you
Once personalization begins, the questions draw on your full history. The same person might receive very different questions depending on what they have written, what patterns are emerging, and what the system has never probed.
You write frequently about decisions you need to make. You rarely write about decisions you have already made and regret. What is one?
You have written about your father and about your work as separate subjects. Is there a connection between them that you have not examined?
What is one small, concrete thing that went right yesterday?
The illustrative questions above are fictional. Real questions are generated from real writing history and are never shown outside of the day they are asked. The point is the structure: the system learns what to ask and when to ask it from the accumulated record of how you write, not just what you write.
Why one question is enough
One question removes the problem of what to write about. The constraint is the gift. You do not stare at a blank page. You respond to a specific prompt that was chosen for you, today, based on everything that came before.
Alice has no dashboard, no settings to tune, no metrics to review. The question is the only surface the system presents to you. All the intelligence, all the accumulated history, all the signal computation resolves into a single sentence that meets you where you are.
After a year of daily writing, no other tool can generate questions this informed. The history is the moat. A new app starts at day one. Alice is at day 365 and the questions show it.
The question is what you experience. Everything else is invisible infrastructure designed to make that question worth answering.