a shared Slack thread where anyone can post examples of GPT assessments they thought were wrong, with a note on why.