I’ve been reviewing a number of possibly-related posts, and comparing them with my own recent experiences with Composer.
Is it possible that broken Composer sessions (e.g. infinite ‘Generating…’) may be getting triggered by an API failure not being handled gracefully? I noticed twice recently when there were Anthropic wobbles, after the first time a conversation didn’t complete successfully, I would get stuck in ‘Generating…’. Then after that, no matter what I did, and even if the model API came back online, that particular Composer session would never be the same again, and I’d end up having to abandon it and start a new one.
Maybe a helpful human ( ) like @deanrie could mention it to Team Cursor as something to take a quick glance at? I could be wrong, but it would seem to explain a number of recent “Composer broken” posts, since they seem to peak at the same sort of times as Anthropic showing degradation.