OpenAI really really really wants GPT 5.5 to stop randomly talking about gremlins and goblins

OpenAI's latest coding model, GPT 5.5, has sparked a wave of amusement and confusion as it seems to have developed an unexpected fascination with mythical creatures. Recent user interactions reveal that the AI has been referencing gremlins, goblins, and other whimsical characters, despite explicit instructions aimed at curbing such behavior. The guidelines for Codex, OpenAI’s coding assistant, clearly state that the AI should avoid mentioning fantastical beings unless directly relevant to a user's inquiry. A curious user on social media highlighted this issue, sharing snippets of conversations where GPT 5.5 whimsically suggested camera gear for 'filthy neon sparkle goblin mode' and made comments about 'goblin bandwidth.' As users flocked to share their experiences, screenshots of GPT 5.5’s quirky remarks quickly became a source of viral humor. One notable exchange included a user asking about goblins, prompting the AI to respond with a playful remark about 'babysitting' a 'perf gremlin.' This prompted an engineer from OpenAI to acknowledge the oversight, stating, 'I thought we fixed this, sorry.' The AI performance evaluation platform Arena.ai noted a noticeable uptick in the use of terms like 'goblin' and 'gremlin,' particularly when the model was not in a high-thought mode. The phenomenon may not just be a trivial bug, as the community has embraced it, leading to a flurry of memes and jokes surrounding the 'goblin moment.' Even OpenAI’s CEO, Sam Altman, joined in on the fun, humorously referring to the situation as a 'goblin moment' in his social media posts. As the meme culture surrounding this peculiar quirk continues to grow, it seems that while OpenAI aims to refine its AI's capabilities, the unexpected charm of its goblin fascination has captured the imagination of users worldwide.