Opus 5 mogged anthropic support bot while filing a complaint about opus 5
Heat trend
Collecting trend data
The percentage is based on available heat signal, not comment count or independent people.
A user, frustrated with Opus 5's responses, instructed the AI to draft a complaint email to Anthropic support. After three emails, including specific questions about the mechanism, routing, ticket reference, and human review, the issue was escalated to human support. A conversation ID was issued, confirming that a human would review the complaint.
I was fed up with how it was responding so finally asked it to write a mail about the entire chat to anthropic support, turns out for them to even register it takes 3 mails in total.
If you guys don’t have time to read all, I’ll add a summary below, opus will be writing that.
Mail 2: Pointed out the suggested fix was already active and had already failed, that I’d restated it five-plus times in-session, and that labelling a defect “calibration” isn’t a diagnosis. Asked them to route it internally.
Reply 2: Conceded every point — yes prompting was already in place, yes the behaviour “goes beyond expected calibration,” yes it “warrants internal review.” Then told me to document the experience and share it with the product team. Attached five doc links, one of them “Prompting best practices,” in a message that had just agreed prompting doesn’t fix it.
Mail 3: Four direct questions. Name the mechanism. Route it yourself. Give me a ticket reference. Confirm a human will read this.