In the early days of Google Chrome, I was tasked with making it work with Windows screen readers. Now, accessibility APIs on Windows were documented, but web browsers used a bunch of additional APIs that were poorly documented. Chrome's design was radically different than Firefox's and IE's, so it was a challenge to implement the APIs correctly. At first I was just trying to get it to work with completely static web pages.

Screen readers were reading form controls, but no matter what I did they weren't activating any of their web-specific features in Chrome. I spent weeks carefully comparing every single API between Firefox and Chrome, making the tiniest changes until they produced identical results - but still no luck.

Finally, out of ideas, I thought to build Chrome but rename the executable to firefox.exe before running it. Surely, I thought, they hadn't hard-coded the executable names of browsers.

But of course they had. Suddenly all of these features started working.

Now that I knew what to ask for, I reached out and made contact with the screen reader vendor and asked them to treat Chrome as a web browser. I learned another important lesson, that I probably shouldn't have waited so long to reach out and ask for help. It ended up being a long journey to make Chrome work well with screen readers, but I'm happy with the end result.