Spending that is seen gets taken seriously
While my car charges I often walk over to the snack shop next door and pick up something to eat. It never felt like much, until the months added up. What follows from that: why a month-end total cannot tell you where money went, why spending that looks reasonable one entry at a time only gives itself away in aggregate, and why the small stuff ended up with an index of its own.
Economic theory inside personal finance
Economic theories I only noticed one by one while building PFEx: whose constraint a budget is, whether pocketing money is a flaw or human nature, what saving is actually for. For each theory, one line on what it says, and one on the design call it became.
Two AI desks: one archives, one acts
On 14 October PFEx gained a second AI role, the Settlement Desk. It and the Research Desk are two very different designs. The Research Desk is not real-time: it files one report whenever a period closes. The Settlement Desk is real-time: you talk to it any time. It grew from a chat box into a private finance agent with skills covering almost every data operation in the app, though it still files a draft and waits for approval before it writes.
Desktop widgets: record straight from the home screen
The entry point for recording a transaction was buried at the bottom of the account detail page. I turned it into an Android home screen widget: one tap opens that account's record sheet, with buttons to switch accounts and a waterline tied to the account's budget.
Redeem codes, and the background service I deleted on purpose
Two things shipped on 9 October: a redeem-code flow, and the removal of always-on background auto-booking. The first exists because mainland China cannot use Google Play, so store billing does not exist for those users; the second was me deciding that notification parsing could not be trusted.
The original concept: one person's finances as one market
PFEx stands for Personal Financial Exchange. Before the first line of code I had already written the whole setup: accounts as securities, balances as prices, net worth as the index, and a full page of things I chose not to build.
Between the idea and the app: a PRD
Before PFEx's first line of code there were documents. This is how I talked an idea through with AI into a PRD, split that PRD into components and milestones, and then built it week by week.