Ever wonder exactly how screenreaders deliver content to users? Confused about how html5 semantic elements, aria-roles and attributes, and good old-fashioned alt text and title attributes and the like interact? Wondering if a seemingly working, straightforward page is bizzarely difficult for screen-reader users or keyboard-only users? Wondering if the "fix" described on Stackoverflow or the Mozilla Developer actually works, in real browsers? Wondering how consistently things work, across browsers?
Well, I am.
Each directory contains a self-contained page. Each experiment builds on the next. Each page contains the markup, a description of what I changed, and a discussion of the results.