

I think this part references it, though it’s kinda solely in passing:
Production evaluations can elicit entirely new forms of misalignment before deployment. More importantly, despite being entirely derived from GPT-5 traffic, our evaluation shows the rise of a novel form of model misalignment in GPT-5.1 – dubbed “Calculator Hacking” internally. This behavior arose from a training-time bug that inadvertently rewarded superficial web-tool use, leading the model to use the browser tool as a calculator while behaving as if it had searched. This ultimately constituted the majority of GPT-5.1’s deceptive behaviors at deployment.


Hi, not OP, but: that’s known as frontmatter, it’s somewhat widespread, and thus I suspect that it’s much more difficult to have it live at the end of your markdown files than in a separate file or db altogether - unless OP is already rolling their own markdown parser.