Financial data and health data are where privacy stops being theoretical
Every previous processes data that is personal but not sensitive in the same way. Your calendar events, email subjects, meeting attendees, journal themes, task lists, and reading highlights are useful . Financial transactions and health measurements are different. They are the most sensitive data in the system, and they require explicit decisions about how they are stored and processed.
This chapter covers two modules (finance and health/energy) together because they share the same design question: where does this data live, and who can see it?
Privacy tiers let you choose how much detail the assistant processes for sensitive data.
The finance radar finds patterns in spending without becoming a budgeting tool
The finance radar detects spending anomalies. You open a export from your banking app, tell the assistant to categorize each transaction, and review the summary it produces. The assistant spots duplicate charges, unusual spikes, and recurring subscriptions you may have forgotten.
'You spent 40 percent more on dining this month. The spike started the week you had four client dinners.' 'You are paying $14.99 per month for a service you last used four months ago.' 'At your current rate, you will have approximately this amount available at month end after known obligations.'
The system notices. You decide what to do. Awareness is the product, not control.
The finance radar surfaces evidence for review; it does not take control of your money.
Category inference improves with the same correction loop as email triage
When you review the categorized list and find a restaurant charge labeled 'Shopping,' correct the category and save the correction as a permanent rule: 'Costco is always Groceries. Venmo payments to this name are Rent.' Each correction improves future categorization. The same feedback loop from the email .
The subscription audit surfaces recurring charges you have forgotten about
A recurring-charge detector surfaces all subscriptions, their costs, and their last-use date. 'You are paying $14.99 per month for this service. You last logged into it four months ago. You are paying $9.99 per month for this service. You last used it six weeks ago.' Over months, the subscription audit can save hundreds of dollars by surfacing charges that have become invisible.
Cross- enrichment connects spending to the rest of your life
A restaurant charge on Tuesday correlates with a client dinner on the calendar. A subscription charge matches a service the reading shows you have not used. An unusual spending spike correlates with a high-stress period in the journal. These cross-references are the OS advantage that no standalone finance app provides.
The energy tracker correlates health data with everything else in the system
You probably already track some health data (a phone step counter, a wearable sleep tracker, or the simple energy rating from the journal ). This module's job is to correlate health data with everything else: calendar density, task completion, journal , email volume.
The minimum viable version uses only the energy rating from the journal: one number per day. Combined with your calendar and task data, that single number reveals patterns. 'Your energy is lowest on days following more than five hours of meetings. Your task completion rate drops by 40 percent on those days.'
For readers who use wearables, periodic imports (weekly or monthly exports from Apple Health, Oura, or similar) add sleep duration, exercise frequency, and heart rate data. The assistant does not need real-time data to find patterns. Monthly imports work fine.
Over months, the system shows you what conditions produce your best days
Over months, you accumulate data on what conditions correlate with good days. The system builds a personalized profile: 'Your best days tend to have seven or more hours of sleep, morning exercise, no more than three meetings, and at least one ninety-minute focus block.'
This is correlation, not causation. The assistant surfaces patterns. You decide what they mean and whether to change anything. 'Days with exercise tend to have higher energy ratings' is useful. 'Exercise causes higher energy' is a claim the data cannot support.
The assistant can show a pattern; your judgment decides what, if anything, it means.
Privacy tiers give you control over what the assistant processes
Financial and health data require an explicit privacy decision. The choices:
You pick the tier. The assistant respects it. For example, you might allow the to see monthly spending totals while keeping individual transaction records restricted to the finance . Most readers start with aggregated access for finance and full access for the energy rating, since a single number per day is low-sensitivity data.
