Follow-up dictations now keep their context: each request carries a conversation_id, so an endpoint that supports it can answer "yes, send it" without starting over.
No release notes available