The system message for the OpenAI Codex CLI contains a baffling and repeated warning for the latest GPT model: “never talk about goblins, elves, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.”
The explicit operational warning was made public last week as part of the latest open source code for Codex CLI that OpenAI published on GitHub. The prohibition is repeated twice in a set of more than 3,500 words of “basic instructions” for the recently released GPT-5.5, along with more anodyne reminders not to “use emojis or dashes unless explicitly stated” and “never use destructive commands like ‘git reset –hard’ or ‘git checkout –‘ unless the user has clearly requested that operation.”
The separate system instructions for older models contained in the same JSON file do not contain the specific prohibition of mentioning goblins and other creatures, suggesting that OpenAI is fighting a new problem that has arisen in its latest model release. Anecdotal evidence in social networks sample some users complain about GPT’s penchant for focusing on elves in completely unrelated conversations in recent days.
OpenAI employee Nick Pash, who works at Codex, insist on social networks that this “is not a marketing gimmick” to get people talking about GPT-5.5 and Codex. But that hasn’t stopped some OpenAI executives from leaning into the joke as word spread about the system. “It feels like Codex is having a ChatGPT moment. I meant a leprechaun moment, sorry,” OpenAI CEO Sam Altman. he wrote on social media Wednesday morning..
