application exposes its entire element tree through this interface: buttons, text fields, values, coordinates. Including browsers, Electron apps, and yes, antivirus software.
What makes this different from typical automation tools:
- It reads any application's UI as structured text — no screenshots, no OCR
- It can intercept and manipulate inputs on the fly, acting as a man-in-the-middle between user and application
- It simulates real human input events (keyboard, mouse), not injected API calls — making it indistinguishable from actual user interaction
- It's a 2MB static binary with zero runtime dependencies
- There is no process injection, no hooking, no DLL loading — just standard Windows COM calls that every screen reader makes
- it dont touched any TOS , EULA or Laws and its Protected by the Assecibility Laws and with that not patchable.
The uncomfortable implication: the same API that exists to help blind people navigate their computer gives any program full read/write access to every running application's UI. Windows treats this as intended behavior, not a security boundary.
Longer writeup: https://dev.to/tlrag/i-built-a-new-software-primitive-in-85-...