Care
A product that criticizes people's work for a living needs a short, enforceable code. This is ours, in the same voice as everything else here.
Headlines, buttons, and paragraphs are fair game. You, your cofounder, your intelligence, your taste, and your effort are not. The model is instructed that the subject of every criticism is a part of the page. If a roast ever gets personal, that is a bug, and we want the report.
No profanity, no slurs, no pile-ons. A good roast is a dry one-liner followed by something useful. If it would not be okay to say to a colleague's face in a page review, it does not belong in the product.
A verdict without a fix is rejected by our schema before it reaches you. The point is a better page by this afternoon, not a bad mood.
We do not publish them, sell them, share them, or train models on them. They go to a model provider through their API for one purpose: producing your roast.
Settings has two buttons. One downloads every roast and every stored page as JSON. The other deletes your account and all of it, immediately. No retention limbo, no exit survey.
Send it to support@bluntpage.com or open a ticket in the help centre. A person reads it, and the prompt or the filters get fixed. One more thing: a low score is a judgment of a page on one day. It says nothing about you or what you are building. Fix the headline and go again.