Case study
Own project · Windows app
AutoDark: Windows theme switching, shipped to Microsoft Store.
Automatic Windows 11 light and dark mode at sunrise and sunset. A focused ON/OFF utility, built by Danilo Stoletovic and published on Microsoft Store.
The application
The problem
I wanted Windows to follow daylight for years. The alternatives I tried brought more complexity and background activity than I wanted. A simple Windows feature shouldn't require complicated software. I wanted a switch, not another hobby.
The solution
I built a native .NET Windows app with a minimal ON/OFF interface. AutoDark calculates sunrise and sunset from Windows location or manually entered coordinates, uses light mode during the day and dark mode at night, and hands scheduled executions to Windows Task Scheduler. Closing the window exits the app; turning automation OFF removes the task and restores the saved theme settings.
Keep the job small
Focused functionality and low background resource use were design constraints. There is no tray icon or resident service: Windows launches brief executions when needed. No account, no telemetry, no settings maze. These are architecture choices, not performance benchmarks.
A package changes the rules
The Windows theme integration had to account for MSIX behavior. The unpackaged app uses narrow registry changes; the current packaged implementation applies a copied theme through Windows Theme Manager, changing the light/dark modes while excluding wallpaper, cursors, icons and sounds from application. That interface is undocumented, so compatibility needs testing rather than assumptions. The scheduled packaged app uses a version-independent execution alias instead of a versioned installation path.
The sandbox said no
Store submission also meant dealing with capability restrictions. A rejected unvirtualizedResources capability was removed. In a separate AppContainer experiment, launching the probe and initializing Theme Manager succeeded, but applying a readable bundled theme was denied and Task Scheduler access failed. That was an investigation, not a shipped sandbox feature. The current production manifest retains runFullTrust for ordinary desktop-user execution; it does not request administrator elevation. These local experiments do not prove behavior on every Windows build or identify the exact package version currently listed in the Store.
Shipping is part of the engineering
I built MSIX packaging and validation around the application's package identity, version, assets and execution alias, with a package prepared for manual Partner Center submission. Listing details, privacy and capability declarations, permission restrictions and compatibility checks all belong to the delivery work. A successful build alone does not get an app into someone's hands.
The result
AutoDark is a working application publicly available through Microsoft Store, under the publisher name Danilo Stoletovic. That is the milestone: a real utility people can install through a public distribution channel. Store publication is distribution, not a Microsoft endorsement.
The button was the easy part
A small interface can hide substantial platform work. Packaging changes execution behavior. Permission models can block the core feature. Distribution has its own requirements. The lesson: shipping software includes compatibility, packaging and getting it to users, not just writing the code.
Check it yourself
Related service: Web apps & AI integrations
My stack
Jack of all trades. Master of none.
Still shipping the whole fucking thing