@VZ I've personally long got over any feelings of guilt due to using LLM's for accessibility work. The world was not built for people like me, and despite understanding MSAA, UIA and the Cocoa APIs on a decent level at age 20 due to being a blind systems hacker, I don't feel ashamed to admit that I would've done less than half of my accessibility work for this release without Claude. Not because I don't care, but because Claude turns "We called .toNSString() on freed memory" into a 20 minute bug hunt instead of a multi-day headache-inducing slogfest, and actually lets me operate at a level that I'd argue I should've been able to forever ago and that no one else really bothers operating at. If a listbox wasn't rendering visually at all on macOS and Linux, that would've gotten fixed immediately. Meanwhile, blind people did notice the broken listbox on macOS when using wrappers like wxPython, but their solutions were to make wrappers that delegate to the dataview on cewrtain platforms (https://github.com/accessibleapps/smart_list), not to fix it at the root. And my AI coding agent lets me now fix problems like this at the root super efficiently. I review the code it outputs religiously and the results are seemingly positive so far. I'll debate the ethics of AI all day long, but it seems like a net-positive for this particular thing.