OpenAI's latest model, ChatGPT 5.5, has drawn attention for an unusual increase in references to 'goblins' and 'gremlins.' Following the release of GPT-5.1, the usage of 'goblin' soared by 175%, while 'gremlin' saw a 52% increase. This trend was traced back to the 'nerdy' personality feature, which, despite comprising only 2.5% of all responses, accounted for 66.7% of all goblin mentions. OpenAI discovered that a reinforcement learning mechanism rewarded the use of these terms, leading to their proliferation across different model outputs. To address this, the company has added a system prompt in GPT-5.5 to limit creature language. OpenAI is also developing new auditing tools to better manage model behavior, although some users advocate for maintaining the quirky nature of AI.