My texts are rubbish because, somehow, the keyboard predictions & autocorrect are worse now than 5 years ago. We have LLMs barfing out fully coherent sentences on their own, how does this even happen.

Didn’t they go through a lot of effort to not have to have many testers?
They have rolling releases and preview channels for regular users to bugtest for them.
All this money can now be spent where it matters - on a new yacht

Hovering army person merging with a tree