Vantage

The problem

Budgeting apps usually show you the past: what you already spent. Vantage's premise is that the useful question is forward-looking, what can I afford this weekend, what will this bill cost over a year, and that answering it honestly means saying how confident the app is.

Constraints

• No account to make, no password, no server holding a copy of your ledger.
• Bank connections are read-only: the app can't pay a bill, transfer money, or cancel anything.
• Transactions are stored only on the phone; the backend is a stateless token broker that never stores financial data.
• Cash App has no official connection any app can make, so that data comes from reading on-device notifications instead.

Approach

Built in SwiftUI with a Node/Express backend that only brokers tokens.

1. Every projected number is paired with the app's own hedge, quoted from the screen it appears on: "this assumes the next few weeks look like the last few," "4 possible subscriptions are not counted."
2. Bank connections stay strictly read-only through Plaid, and the backend never stores a transaction.
3. For Cash App, which has no API, the app reads notifications on-device and parses them into transactions; manual import or typed-in entry is the fallback for any account.

What was hard

Cash App's notification text had more shapes than expected: settling a gas station hold from its "authorization updated" wording, writing the cash-out side so the balance comes down once money lands at the bank with the instant-deposit fee booked as its own line, and handling cases where iOS delivers two notifications as one. Spanish-language notification parsing came later, in the same pass that fixed three gaps a separate audit had found.

Result

In testing, invite only. Plaid runs in production, and a screenshot tour runs on iPhone and iPad simulators as part of the test suite.