berndt.codes

hefted

Status
iOS · beta
Role
Idea · UX · Development · Operations
hefted preview

There is always a stack of paper lying around at my place that is important. Or maybe not, you never quite know. hefted dissolves exactly that stack: photograph the paperwork, and the app names it, files it, and stores it so you can find it again, so that you are then allowed to forget it. The AI that sorts it runs entirely on the device, with no account and no server.

Where it started

hefted came out of a frustration: I keep forgetting to file receipts in a way that lets me find them again. The reason is always the same, it takes effort. Sometimes you have the receipt on paper but no scanner app to hand, sometimes it comes by email or through Amazon, sometimes it is pressed into your hand at the checkout. Every time you have to remember to digitize it, file it somewhere, and pick a filename that lets you find it again for a warranty or your taxes. And even when I kept it up for a while, in the end a file still landed in some folder that I meant to tidy up later and then never did.

This is not only about receipts. Vouchers arrive by post, in a flyer, or as a gift, and then they sit at home and you never have them with you when you need them. The same goes for letters and everything else you think you cannot throw away, in case you still need it. In the end it lands on a stack of paper, until you eventually work through it.

Photograph it and be allowed to forget

The usability of hefted has exactly one job: to take the worry away. You start the camera scanner straight from the Action Button, photograph the document, and put the phone away again. The app handles the rest in the background, as long as the system lets it, for instance while the device is charging, and entirely without internet. You do not have to think about the filename, where it is stored, or finding it again. You are simply allowed to forget it until you need it.

The AI stays on the device

The trigger was the moment when, with Apple Intelligence, an AI became available directly on the iPhone. Exactly the dull work of naming a document and recognizing what on it matters is something it can take over. That convinced me to build the whole thing consistently on the device. The documents do not leave the phone, there is no account and no server, and where the files end up is for the user to decide, from an iCloud folder to their own NAS. This is not only a privacy decision. It makes the app usable offline and keeps it lean.

Built without a single line of Swift

I built hefted entirely with Claude Code, and I do not write Swift. That was the real draw of the project: to find out whether I can build an app with AI tools alone without ending up with junk, but with something solid and secure.

The key was not to code myself, but to lead the development. I planned, set the architecture, laid down coding rules, checked every change, and insisted on tests. So exactly the role I also take in a team, except that the team here was an AI tool. Idea, design, UX, and operation are mine anyway. What came out is a native iOS app whose quality I stand behind, in a language I do not write myself.

Results

  • Scan it, the app does the restScan a document via the Action Button, and the app names it, files it, and stores it.
  • AI stays on deviceThe documents never leave the iPhone.
  • No account, no serverFree choice of storage, from an iCloud folder to your own NAS.
  • Offline, one-time purchaseWorks without internet, with a tax export and a one-time purchase instead of a subscription.

Built with

  • Swift
  • SwiftUI
  • Apple Intelligence (on-device)
  • Claude Code

Back to work