@aral I just see AT as "protocol competition" with AP. My main concern is why it didn't go through W3C. Optimists will say "to move faster", pessimists will say "they wanna monopolise". It's probably both.
Despite this, the spec itself seems to be good and solve some (all?) of AP's issues. I would love to see an expert, like @evan go in depth about AT's flaws, as a protocol. I'm sure it has some.