I gave three assistants a commentary that was 0.2 short. Two said it was fully explained.
Four month-end prompts for finance teams, each run on three assistants before sending, and the one practice under all of them: the assistant drafts, a person approves.
I gave three assistants a status update. Two added a deadline problem that was not in it.
Four prompts that make an assistant declare what it did to your text. I ran all four before sending this, and the results are less flattering than I wanted.