QNAHow it works
From a screen you are already looking at, to an answer.
QNA runs in the tray on Windows and reads the display the machine already has. These are the five things it does, in the order a question meets them.
Watching
The picture comes from the display, not of it
A camera exists to turn a screen into pixels, and Windows can hand those over directly. QNA reads the display through the system itself, so the capture is square-on and as sharp as what you see — there is no perspective to correct and no glare to work around.
Choosing
Capturing and answering are separate jobs
Capture only has to keep up with a scrolling page, so it stays cheap. Everything expensive happens afterwards, over a handful of frames chosen on this machine. That is what lets one pass down a long list turn into a set of answers without answering the same question six times on the way.
Repairing
A question cut in half is put back together
Scroll through a list and some questions are split: the top of one frame carries what the bottom of the last one cut off. QNA joins the halves before answering, so what it answers is the whole question rather than the part that happened to fit.
Answering
Once per question, and remembered
Each question is answered once. The answer is kept with the picture it was read from, so you can see what it actually looked at, and answering the same question again updates what is there rather than adding another copy.
Serving
One page, and it is the controls
QNA serves a page on this machine, laid out for a phone as well as a desktop. Start, stop, a single capture and a scan are all on it, so the phone in your hand drives the Windows machine rather than just reporting on it.
Two ways in
The same page, from the desk or from the sofa.
On the machine itself the page is at localhost. From a phone it is at the machine's own address on your network, which About prints for you. Either one can start a scan.