ScreenSense

ScreenSense starts with a reasonable question: if I am already looking at something on the phone, can software use that context to help with what I am doing?

The reasonable question acquires Android permissions, capture lifecycles and background-service behaviour almost immediately.

The project explores screen capture, accessibility information and text extraction. Those are different ways of finding useful context. A structured interface can expose text directly; other screens leave only pixels to work with. The implementation needs to account for both without treating the display as an unlimited stream of available data.

Capture is a capability, not blanket access

I want capture to be deliberate and understandable. The useful boundary is being able to explain what is captured, what processes it and when that access ends. A service restarting should not quietly be treated as proof that an earlier permission still applies.

The project is still experimental. Code and build notes establish what has been implemented; claims about reliable operation on a particular phone need device evidence. Source code containing a capture path is not the same as that path surviving every interruption the operating system can throw at it.

ScreenSense sits beside the broader companion work as a focused investigation into screen context. Its job is to make the information useful while preserving the distinction between what the system can inspect and what it should act on.