This article from Salma really resonates with me and honestly explains why I'm so sour about AI code assistant tools. Anytime I scoff at using them, inevitably someone says something along the lines of "But they're coming! They're all getting better!" Sure, I'll wait for that day then.
I am eager for the days where I'm not spending time reminding people to add their missing alt text, or to convert a div to a button where appropriate. The absolute basics. AI is not solving those problems for us. It's either making them worse or just doing them in weirder ways.
Before, I would help engineers of all experience levels learn why their ARIA attributes were misused or why their code would result in a poor experience for keyboard users. Sure, they could have copy and pasted code from Stackoverflow. But usually, there was an intentional act and some scrutiny that came from them, where they could explain why they added that code. This leaves you with something to learn from. Now, from engineers that use these tools, I fear there is less scrutiny and less reflecting on their decisions with code. They will ultimately learn less.
These tools are also not affording me time to write the fun parts of code. Nor are they enabling what I think is truly the driving force behind why I care about web accessibility: that accessibility is at the root of more creative and more-well-loved software and tooling, for everyone. If anything, I find these AI tools are making less space for it. My advocacy is now not just "Here is why you should make accessible software" but ALSO "Here are the dangers of using AI tools to write your software". When do I get to move onto the more creative and fulfilling parts of working in web accessibility? As Salma says, that promise wasn't kept.