The file system is nice because it's the same database. I don't want all my programs to do their own database, since then how do I search and move stuff in bulk?

No, but all your programs can/could share the same database, and you could cross-reference anything whenever needed. At least that's how I manage it. I have a single postgresql database with per-program/per-instance schemas.

I just did my taxes and figuring out what to pay was just a single select over some tables in the paypal and various per-bank schemas + a schema that had a table of currency conversion rates from our central bank applicable for each month of the year for tax purposes, expected by the tax man (or woman). Quick and easy. I don't even bother with UI, for these once or twice a year needs, just like I wouldn't bother with writing UI for some mp3 conversion task. Just a simple script will do.

Filesystem is great for arbitrary data/files with no schema. Random pdf files, code, programs, etc. But anything that has some obvious schema and comes in large quantities and perhaps needs to be modified/synced with third party data source, I like having such things in the database. It's so much more useful that way, because it's much easier to do something with the data.

A local database will store data in a file, but I get what you're saying. Files are a common interface that allow you to pipe data around, keep it portable and malleable.