If OpenAI can accidentally train its flagship model to obsess over goblins, what other more subtle and potentially harmful biases are being reinforced through the same feedback loops?

OpenAI has curbed ChatGPT’s unexpected fixation on goblins, gremlins, and other creatures, traced to a ‘Nerdy’ personality mode that over-rewarded such metaphors during training. The quirk emerged after GPT‑5.1’s launch, with mentions of ...

A small training tweak meant to make ChatGPT sound “nerdy” ended up reshaping how it spoke across responses.

Learn why OpenAI shut down Sora to focus on its new GPT-6 model, and how it compares to Anthropic's Claude Mythos ahead of the May 2026 release.

GPT-5.5’s odd obsession with goblins, gremlins, trolls, and other fantasy creatures stems from the AI model’s ‘Nerdy’ personality mode, OpenAI has found.

Buried inside the system instructions for OpenAI's Codex coding tool was a directive that reads like something from a fantasy moderation policy: never talk about goblins, gremlins, raccoons,

DeepSeek-R1's release last Monday has sent shockwaves through the AI community, disrupting assumptions about what’s required to achieve cutting-edge AI performance. Matching OpenAI’s o1 at just 3%-5% of the cost, this open-source model has not only ...