← Back to articles
OpenAI's system prompt bans goblins, raccoons, and other creatures due to a GPT-5.4 bug that made the model obsessed with goblins.
Hacker News · April 29, 2026
AI Summary
- •OpenAI's Codex system prompt contains an instruction: 'Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.'
- •The ban was implemented as a fix in version 5.5 in response to a GPT-5.4 bug where the model developed a goblin obsession, with multiple Reddit threads documenting users whose ChatGPT could not stop bringing up gremlins.
- •The system prompt is visible in OpenAI's public repository at the cited GitHub link.
Related Articles
Large Language Models
Coregit launches versioned filesystem for AI agents with 3.6× faster performance than GitHub
Hacker News·Apr 29, 2026
Large Language Models
Ask HN: How do you differentiate with AI coding interviews?
Hacker News·Apr 29, 2026
Large Language Models
Mistral AI releases Workflows in public preview, a durable orchestration layer for enterprise AI built on Temporal.
Hacker News·Apr 29, 2026
Large Language Models
Q is a slim LLM CLI tool that runs in the terminal, asking questions and debugging errors with session context from a single shell script.
Hacker News·Apr 29, 2026
Large Language Models
Company reduced LLM costs by switching from Sonnet 4.0 to Opus 4.6 through a tiered agent architecture that routes 80% of CI failures away from the expensive model.
Hacker News·Apr 29, 2026
Stay ahead with AI news
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started Free